Next Generation Sequencing of the Upper Respiratory Tract Microbiota

The microbiome of otitis media and development of a probiotic to prevent otitis media in Indigenous Australian children

Andrea Coleman Doctorate of Medicine; Bachelor of Speech Pathology (Hons I)

https://orcid.org/0000-0001-8101-1585

A thesis submitted for the degree of Doctor of Philosophy at The University of Queensland in 2020

Faculty of Medicine 1

Abstract Background Indigenous Australian children have endemic rates of otitis media (OM), impacting negatively on development, schooling and employment. Current attempts to prevent and treat OM are largely ineffective. Beneficial microbes are used successfully in a range of diseases and show promise in OM in non-Indigenous children. We aim to explore the role of beneficial microbes in OM in Indigenous Australian children.

Aims 1) Explore the knowledge gaps pertaining to upper respiratory tract (URT)/ middle ear microbiota (pathogens and commensals) in relation to OM in indigenous populations globally by systematic review of the literature. 2) To explore the URT microbiota in Indigenous Australian children in relation to ear/ URT health and infection. 3) To explore the ability of commensal bacteria found in the URT of Indigenous children to inhibit the growth of the main otopathogens.

Methods The systematic review of the PubMed database was performed according to PRISMA guidelines, including screening of articles meeting inclusion criteria by two independent reviewers.

To explore the URT microbiota, we cross-sectionally recruited Indigenous Australian children from two diverse communities. Demographic and clinical data were obtained from parent/carer interview and the child’s medical record. Swabs were obtained from the nasal cavity, buccal mucosa and palatine tonsils and the ears, nose and throat were examined. Samples were analysed using a culturomics approach with MALDI-TOF isolate identification. Real-time PCR was used to qualify otopathogen loads and detect respiratory viruses. Culture-independent analysis of the nasal microbiota was examined using16S rRNA amplicon next generation sequencing (NGS).

The bacterial interference of lactobacilli and alpha haemolytic streptococci (AHS) were investigated using agar overlay and cell-free supernatant. Promising isolates underwent whole genome sequencing (WGS) to investigate genetic markers of URT tropism, antibiotic resistance and

virulence genes. In vitro antibiotic susceptibility was examined for ampicillin, amoxicillin + clavulanic acid, and azithromycin.

Results The systematic review included 25 papers encompassing Indigenous Australian, Alaskan, and Greenlandic children. It identified high rates of nasopharyngeal colonisation with the main otopathogens in indigenous children with OM. There was significant heterogeneity between studies, particularly in microbiological methods, which were largely limited to culture-based detection of the main otopathogens with an absence of data regarding commensal bacterial flora in the upper airway.

We recruited 103 Indigenous Australian children aged 2-7 years (mean 4.7 years). Seventeen (16.5%) children were ‘healthy’ (normal examination and no history of OM). Investigation of nasal microbiota showed that children with a history of OM/ current OM/URT infection (URTI) had higher otopathogen detection and loads, and rhinovirus detection compared to healthy children (all p < 0.04). Investigation of network relationships revealed a strong correlation between high otopathogens loads in children with a history of OM/ current OM/URTI. Healthy children demonstrated a more complex network of correlated genera and a strong correlation between Corynebacterium pseudodiphtheriticum and Dolosigranulum pigrum. 16S NGS showed that Dolosigranulum was ubiquitous across all otitis groups but correlated with different genera in each group. Ornithobacterium was only detected with 16S NGS and was identified in children with current/ historical OM. It was absent/ at low relative abundance in the healthy children. Ornithobacterium was strongly correlated with Helcococcus, Dichelobacter and clustered around Streptococcus and Haemophilus.

In relation to nose health, children with purulent rhinorrhoea had higher nasal otopathogen detection and loads, and rhinovirus detection compared to those with healthy noses (all p < 0.04). Children with healthy noses had a strong correlation between C. pseudodiphtheriticum and D. pigrum.

Twenty-six lactobacilli isolates and 66 AHS isolates from 17 remote children were tested against otopathogens. Lactobacilli could readily inhibit the growth of otopathogens; three Lactobacillus rhamnosus isolates were more effective than commercially available strains, L. rhamnosus GG and

L. rhamnosus LB21. AHS were less effective inhibitors, although some isolates were able to inhibit Streptococcus pneumoniae. Three lactobacilli progressed to WGS. One, L. rhamnosus (3160), had SpaCBA genes coding for pili to adhere to epithelial cells. We detected minor antibiotic resistance genes coding for antibiotic efflux pump and a ribosomal protection protein, neither associated with typical URT antimicrobials. The lactobacilli were susceptible to typical URT antimicrobials in vitro. Screening for virulence genes detected genes for two putative capsule proteins that have been described in bacteria from other genera.

Conclusion We have demonstrated the importance of bacterial relationships in the expression of URT health or disease. Poor ear/ URT health is related to strong correlation between high otopathogens loads, suggesting otopathogen synergism. Healthy children demonstrate a strong relationship between C. pseudodiphtheriticum and D. pigrum, which is not seen in other phenotypes, suggesting that C. pseudodiphtheriticum-D. pigrum synergism supports URT health. We detected Ornithobacterium, likely Candidatus Ornithobacterium hominis, and in this population was correlated with a novel bacterium which appears to be related to poor upper respiratory tract/ear health. We found lactobacilli that readily inhibited otopathogens with in silico and in vitro support a positive safety profile.

Declaration by author

This thesis is composed of my original work, and contains no material previously published or written by another person except where due reference has been made in the text. I have clearly stated the contribution by others to jointly-authored works that I have included in my thesis.

I have clearly stated the contribution of others to my thesis as a whole, including statistical assistance, survey design, data analysis, significant technical procedures, professional editorial advice, financial support and any other original research work used or reported in my thesis. The content of my thesis is the result of work I have carried out since the commencement of my higher degree by research candidature and does not include a substantial part of work that has been submitted to qualify for the award of any other degree or diploma in any university or other tertiary institution. I have clearly stated which parts of my thesis, if any, have been submitted to qualify for another award.

I acknowledge that an electronic copy of my thesis must be lodged with the University Library and, subject to the policy and procedures of The University of Queensland, the thesis be made available for research and study in accordance with the Copyright Act 1968 unless a period of embargo has been approved by the Dean of the Graduate School.

I acknowledge that copyright of all material contained in my thesis resides with the copyright holder(s) of that material. Where appropriate I have obtained copyright permission from the copyright holder to reproduce material in this thesis and have sought permission from co-authors for any jointly authored works included in the thesis.

Publications included in this thesis

Coleman A, Wood A, Bialasiewicz S, Ware RS, Marsh R. L., & Cervin, A. (2018). The unsolved problem of otitis media in indigenous populations: A systematic review of upper respiratory and middle ear microbiology in indigenous children with otitis media. Microbiome, 6(1), 1–15.

Coleman A, Bialasiewicz S, Marsh RL, Grahn Håkansson E, Cottrell K, Wood A, Jayasundara N, Ware RS, Zaugg J, Sidjabat HE, Adams J, Ferguson J, Brown M, Roos K, Cervin A. Upper respiratory microbiota in relation to ear and nose health among Australian Aboriginal and Torres Strait Islander children. J Pediatric Infect Dis Soc. In press.

Submitted manuscripts included in this thesis Coleman A, Zaugg, J, Wood A, Cottrell K, Grahn Håkansson E, Adams J, Brown M, Cervin A, Bialasiewicz S. (2020). The upper respiratory tract microbiome of Australian Aboriginal and Torres Strait Islander children in ear and nose health and disease; a prospective cohort study. Under Review.

Coleman A, Håkansson A, Grahn Håkansson E, Bialasiewicz S, Zaugg J, Cervin, A. (2020). Inhibition of respiratory pathogens by lactobacillus and alpha haemolytic streptococci from Aboriginal and Torres Strait Islander children. Journal of Applied Microbiology. Under Review.

Other publications during candidature

Peer-reviewed Papers: Coleman A & Cervin A. (2019). Probiotics in the treatment of otitis media. The past, the present and the future. International Journal of Pediatric Otorhinolaryngology, 116, 135–140

Conference Abstracts Coleman A. et al. Microbiome of the upper respiratory tract in Australian Indigenous children. The Australian Society of Otolaryngology Head and Neck Surgery’s Annual Scientific Meeting. 2019 Coleman, A et al. Is Dolosigranulum a potential microbiome therapeutic for otitis media? Otitis Media Australia Conference (OMOZ). 2018. Coleman A, et al. The upper respiratory tract microbiota in relation to otitis media in 2 contrasting Australian Indigenous communities: Implications of generalisation. The European Society of Paediatric Otolaryngology and Otitis Media Australia Conference (OMOZ). 2018 Coleman A, Wood A, Bialasiewicz S, Cervin. The unsolved problem of otitis media in Indigenous populations, is the answer to be found in the commensal flora? 4th South Pacific Otorhinolaryngology Forum. 2017 Sidjabat H, Coleman A, et al. Commensal bacteria of the upper respiratory tract in relation to otitis media in remote Indigenous Australian children. 4th South Pacific Otorhinolaryngology Forum. 2017

Contributions by others to the thesis Professor Anders Cervin, Dr Seweryn Bialasiewicz, and Dr Eva Grahn Håkansson provided significant assistance in the design, execution, analysis and interpretation of studies within this thesis. Dr Grahn Håkansson, Dr Bialasiewicz, Kyra Cottrell, Alexander Håkansson, and Nadeesha Jayasundara provided assistance with laboratory work. Amanda Wood assisted with sample collection and community consultation. Professor Robert Ware, Dr Robyn Marsh, and Dr Julian Zaugg provided substantial assistance with statistical analysis and bioinformatic design, execution and interpretation.

Statement of parts of the thesis submitted to qualify for the award of another degree “No works submitted towards another degree have been included in this thesis”.

Research Involving Human or Animal Subjects Ethics approval was obtained from the following Human Research Ethics Committees: • Far North Queensland Human Research Ethics Committee; HREC Reference number: HREC/15/QCH/10 – 954 • The University of Queensland Institutional Research Ethics; Approval Number 2016000292

Acknowledgements I would like to acknowledge Queensland Health’s Deadly Ears team, including Matthew Brown, Amanda Wood, Josephine Ferguson and Jasmyn Adams for their guidance and support in community consultation and engagement and sample collection. I would also like to acknowledge the valuable contributions of Jason Leon, Chantel Hunter, Gail Wason and Deborah Gertz from Mulungu Aboriginal Corporation Medical Centre, Isabel Toby from Save the Children, Doomadgee and Anne O’Keefe from Doomadgee Community Health.

Financial support This research received the following financial support: • National Health and Medical Research Council Postgraduate Scholarship • Avant Doctors in Training Research Scholarship - recipient in the Advancement of Medicine category • Queensland Health Junior Doctor Research Fellowship

Keywords Maximum 10 words; Indigenous Australian, otitis media, nose, microbiota, respiratory virus, otopathogen, probiotic, bacterial interference

Australian and New Zealand Standard Research Classifications (ANZSRC)

ANZSRC code: 920109, Infectious Diseases 30% ANZSRC code: 111403, Paediatrics 20% ANZSRC code: 060501 Bacteriology 30% ANZSRC code: 060506 Virology 10%

Fields of Research (FoR) Classification

FoR code: 1114, Paediatrics and productive medicine 50% FoR code: 0605, Microbiology, 50%

Table of Contents

ABSTRACT ...... 2

DECLARATION BY AUTHOR ...... 5

PUBLICATIONS INCLUDED IN THIS THESIS ...... 6

SUBMITTED MANUSCRIPTS INCLUDED IN THIS THESIS ...... 7

OTHER PUBLICATIONS DURING CANDIDATURE ...... 7

PEER-REVIEWED PAPERS: ...... 7

CONFERENCE ABSTRACTS ...... 7

CONTRIBUTIONS BY OTHERS TO THE THESIS ...... 8

STATEMENT OF PARTS OF THE THESIS SUBMITTED TO QUALIFY FOR THE AWARD OF ANOTHER DEGREE ...... 9

RESEARCH INVOLVING HUMAN OR ANIMAL SUBJECTS ...... 9

ACKNOWLEDGEMENTS ...... 10

FINANCIAL SUPPORT ...... 10

KEYWORDS ...... 10

AUSTRALIAN AND NEW ZEALAND STANDARD RESEARCH CLASSIFICATIONS (ANZSRC) ...... 11

FIELDS OF RESEARCH (FOR) CLASSIFICATION ...... 11

TABLE OF CONTENTS ...... 12

LIST OF TABLES ...... 17

SUPPLEMENTARY TABLES ...... 18

LIST OF FIGURES ...... 19

SUPPLEMENTARY FIGURES ...... 21

LIST OF ABBREVIATIONS USED IN THE THESIS ...... 22

CHAPTER 1: INTRODUCTION ...... 24

OTITIS MEDIA ...... 24 1.1.1 Definitions ...... 24 1.1.2 Epidemiology and cost of otitis media ...... 24 1.1.3 Pathogenesis of otitis media ...... 25 1.1.4 Risk factors for otitis media ...... 26 1.1.5 Current treatment and prevention of otitis media in Indigenous Australian children ...... 27 1.1.6 Complications and consequences of otitis media ...... 28 12

UPPER RESPIRATORY TRACT MICROBIOTA ...... 29 1.2.1 Methods to analyse the upper respiratory tract microbiota ...... 29 1.2.2 Development of the nasopharyngeal microbiota ...... 30 1.2.3 Early nasopharyngeal microbiota in Australian Indigenous children ...... 33 1.2.4 Development of mucosal immunity ...... 33 1.2.5 Factors That Disrupt the Nasopharyngeal Microbiota ...... 34

UPPER RESPIRATORY AND MIDDLE EAR MICROBIOTA IN RELATION TO OTITIS MEDIA ...... 37 1.3.1 Bacteriology of otitis media ...... 37 1.3.2 Virology of otitis media ...... 41 1.3.3 Biofilm and intracellular bacteria in otitis media ...... 42

BACTERIAL INTERFERENCE IN THE UPPER RESPIRATORY TRACT ...... 43

PROBIOTICS: SUCCESSES IN MEDICINE ...... 44

PROBIOTICS IN THE UPPER RESPIRATORY TRACT ...... 45 1.6.1 Local administration of probiotics in OM ...... 46 1.6.2 Systemic administration of probiotics in OM ...... 50

NON-PROBIOTIC COMMENSALS TO PREVENT OTITIS MEDIA ...... 55

CONCLUSION ...... 55

AIMS AND HYPOTHESES ...... 57 1.9.1 Aim 1: To explore the current microbial data in relation to otitis media in indigenous populations around the world ...... 57 1.9.2 Aim 2: To identify health-associated bacterial species in the upper respiratory tract of Indigenous Australian children ...... 58 1.9.3 Aim 3: To explore the ability of health-associated bacterial species to inhibit the growth of otopathogens 58 1.9.4 Aim 4: To complete first stage testing for probiotic development for bacterial strains identified as potential probiotic candidates in Aim 3.1 ...... 59

CHAPTER 2: SYSTEMATIC REVIEW ...... 60

2.1 INTRODUCTION ...... 64

2.2 METHODS ...... 65 2.2.1 Inclusion criteria ...... 65 2.2.2 Search strategy ...... 65 2.2.3 Data extraction ...... 66 2.2.4 Data analysis ...... 66 2.2.5 Risk of bias assessment ...... 66

2.3 RESULTS ...... 67 2.3.1 Risk of bias assessment ...... 67 2.3.2 Heterogeneity ...... 69 13

2.3.3 OM clinical definitions and diagnosis ...... 76 2.3.4 Laboratory methods ...... 76

2.4 BACTERIOLOGY ...... 76 2.4.1 Acute otitis media ...... 76 2.4.2 Otitis media with effusion ...... 79 2.4.3 Chronic suppurative otitis media ...... 81 2.4.4 Nasopharyngeal carriage as a risk factor for otitis media ...... 83 2.4.5 Virology ...... 83 2.4.6 Biofilm ...... 83

2.5 DISCUSSION ...... 84 2.5.1 Limitations of the current literature ...... 85 2.5.2 Future directions ...... 86 2.5.3 Conclusions ...... 87

2.6 SUPPLEMENTARY MATERIALS ...... 88 2.6.1 Supplementary Data: Search Strategy ...... 88

CHAPTER 3: COHORT STUDY DESIGN ...... 101

COMMUNITY RECRUITMENT ...... 101

COMMUNITY CONSULTATION AND ENGAGEMENT ...... 101

PARTICIPATORY ACTION RESEARCH ...... 102

COMMUNITY ENGAGEMENT AND INFORMED CONSENT ...... 103

PARTICIPANTS ...... 103

INCLUSION CRITERIA ...... 104 3.6.1 Exclusion criteria ...... 104 3.6.2 Sample size calculation ...... 104

ETHICS ...... 104

RECRUITMENT ...... 104

DATA COLLECTION ...... 105 3.9.1 Demographic data ...... 105 3.9.2 Clinical examination ...... 106 3.9.3 Swab collection and transportation ...... 106

COMMUNITY CONSULTATION ON PROBIOTIC INTERVENTION ...... 108

REFLECTIONS ON LIMITATIONS AND CHALLENGES ...... 108

CHAPTER 4: UPPER RESPIRATORY TRACT MICROBIOTA IN RELATION TO OTITIS MEDIA, UPPER RESPIRATORY TRACT HEALTH AND DEMOGRAPHIC VARIABLES: CULTUROMICS AND PCR ...... 110

4.1 BACKGROUND ...... 115

4.2 METHODS ...... 116 14

Study Design and Sample Collection ...... 116 Microbiological assays ...... 117 Statistical analysis ...... 118

4.3 RESULTS ...... 119 4.3.1 Microbiome across anatomical sites...... 121 4.3.2 Nasal microbiota in relation to ear health ...... 125 4.3.3 Nasal microbiota in relation to nasal health ...... 126 4.3.4 Differences in nasal microbiota across geographic regions ...... 129

4.4 DISCUSSION ...... 129

4.5 CONCLUSION ...... 132

4.6 ACKNOWLEDGMENTS: ...... 133

4.7 SUPPLEMENTARY MATERIAL ...... 134 4.7.1 Supplementary methods ...... 134 4.7.2 Supplementary Tables and Figures ...... 138

CHAPTER 5: NEXT GENERATION SEQUENCING OF THE UPPER RESPIRATORY TRACT MICROBIOTA ...... 156

INTRODUCTION ...... 161

METHODS ...... 161 Population and sample collection ...... 161 DNA extraction and quality assurance ...... 162 16S Sequencing ...... 162 Culturomic analysis ...... 163

RESULTS ...... 163 5.3.1 Nasal Microbiota in relation to ear health ...... 164 5.3.2 Nasal microbiota in relation to nose health ...... 169 5.3.3 Nasal microbiota in relation to season, household occupancy and community ...... 171

DISCUSSION ...... 172

SUPPLEMENTARY MATERIAL ...... 176 1.5.1 Supplementary methods ...... 176 1.5.2 Supplementary Tables and Figures ...... 179

CHAPTER 6: BACTERIAL INTERFERENCE ...... 188

6.1 INTRODUCTION ...... 191

6.2 MATERIALS AND METHODS: ...... 192 6.2.1 Participants ...... 192 6.2.2 Bacterial interference ...... 193 6.2.3 Antibiotic Susceptibility ...... 194 6.2.4 Whole genome sequencing ...... 195 15

6.2.5 Statistical analysis ...... 195

6.3 RESULTS ...... 196 6.3.1 Agar overlay ...... 196 6.3.2 Cell-free supernatant ...... 199 6.3.3 Case Series ...... 200 6.3.4 Whole genome sequencing analysis ...... 201

6.4 DISCUSSION ...... 202

6.5 ACKNOWLEDGEMENTS: ...... 206

6.6 SUPPLEMENTARY DATA ...... 207 6.6.1 Supplementary methods ...... 207 6.6.2 Supplementary Tables and Figures ...... 209

CHAPTER 7: GENERAL DISCUSSION ...... 227

7.1 HEALTHY INDIGENOUS CHILDREN HAD UNIQUE NASAL MICROBIOTA CHARACTERISED BY A RELATIONSHIP BETWEEN C.

PSEUDODIPHTHERITICUM AND D. PIGRUM ...... 227

7.2 OTOPATHOGEN PRESENCE, LOAD AND CO-OCCURRENCE RELATES TO OTITIS MEDIA AND RHINORRHOEA ...... 228

7.3 RHINOVIRUS WAS RELATED TO OTITIS MEDIA AND RHINORRHOEA ...... 229

7.4 ORNITHOBACTERIUM MAY BE A NOVEL PATHOGEN IN THE NOSE OF INDIGENOUS AUSTRALIAN CHILDREN ...... 229

7.5 REMOTE CHILDREN HAD HIGHER OTOPATHOGEN PREVALENCE AND LOAD COMPARED TO RURAL CHILDREN ...... 230

7.6 LACTOBACILLUS FROM HEALTHY INDIGENOUS AUSTRALIAN CHILDREN ARE POTENT INHIBITORS OF OTOPATHOGENS WHILE AHS

WERE POOR INHIBITORS, PARTICULARLY OF M. CATARRHALIS AND H. INFLUENZAE ...... 231

7.7 AHS FROM HEALTHY INDIGENOUS AUSTRALIAN CHILDREN WERE MORE EFFECTIVE AT INHIBITING THE GROWTH OF

OTOPATHOGENS COMPARED TO AHS FROM CHILDREN WITH CSOM ...... 232

7.8 PROMISING LACTOBACILLI HAD LOCI FOR PILI, WERE SUSCEPTIBLE TO URT ANTIMICROBIALS AND HAD NO SIGNIFICANT ANTIBIOTIC

RESISTANCE OR VIRULENCE GENES ...... 233

7.9 STRENGTHS AND LIMITATIONS ...... 233

7.10 FUTURE DIRECTION ...... 235

7.11 CONCLUSIONS ...... 236

REFERENCES ...... 237

APPENDIX I: CLINICAL RECORD FORM ...... 260

List of Tables

Table 1: Definitions of otitis media (Kong et al ((7)) ...... 24 Table 2: Treatment guidelines for Indigenous Australian children (5) ...... 27 Table 3: Randomized controlled trials investigating local (intranasal) administration of probiotics to prevent/ treat otitis media ...... 48 Table 4: Randomized controlled trials investigating systemic (oral) administration of probiotics to prevent/ treat otitis media ...... 53 Table 5: Novel commensal species in development to prevent infectious diseases, from Marsh et al (41) ...... 55 Table 6: Risk of bias assessment ...... 68 Table 7: Characteristics of included studies ...... 70 Table 8: Clinical Definitions ...... 117 Table 9: Demographic and clinical details of participants ...... 120 Table 10: Demographic and clinical details of participants ...... 164 Table 11: Clinical and demographic details of sub-group participants (n = 17) ...... 196 Table 12: Ratio of lactobacillus and alpha haemolytic streptococci (AHS) able to completely inhibit the growth of respiratory pathogens, all isolated from the same child, to the total number of lactobacillus or AHS isolated from that child; and the total number of lactobacillus and AHS having no inhibitory activity...... 200

Supplementary Tables Supplementary Table 1: Summary of OM diagnostic criteria used in studies ...... 89 Supplementary Table 2: Summary of three main otopathogens in indigenous children with otitis media ...... 90 Supplementary Table 3: Summary of microorganisms reported by studies of URT and/or middle ear specimens using specialist laboratory methods ...... 94 Supplementary Table 4: Summary of microorganisms identified in the nasopharynx/ middle ear using next generation sequencing ...... 99 Supplementary Table 5: Species specific-PCR oligonucleotide sequences ...... 138 Supplementary Table 6: Otopathogen load inter-quartile ranges (genome equivalents/µl of nucleic acid extract) ...... 143 Supplementary Table 7: Detection of otopathogens in 101 nasal samples using culture and PCR . 143 Supplementary Table 8: Prevalence of bacterial species in relation to anatomical site ...... 144 Supplementary Table 9: Otopathogen detection according to clinical status ...... 153 Supplementary Table 10: Otopathogen loads (genome equivalents/µl of nucleic acid extract) according to clinical variables ...... 154 Supplementary Table 11: Significantly differentially abundant genera according to DESeq2 ...... 179 Supplementary Table 12: Significantly differentially abundant Dolosigranulum ASVs in relation to otitis status and nose health ...... 181 Supplementary Table 13: Differences in alpha diversity according to otitis status, nose health, community, household occupancy, and season of collection ...... 183 Supplementary Table 14: Species included in agar overlay bacterial interference ...... 209 Supplementary Table 15: Lactobacillus (n = 26) used in bacterial interference studies ...... 210 Supplementary Table 16: Alpha haemolytic streptococci (n = 65) used in bacterial interference studies ...... 212 Supplementary Table 17: Virulence genes with percentage identification >50% ...... 215

List of Figures Figure 1: Proposed model to explain chronic OM in Indigenous Australian children (Wiertsema & Leach, 2009 (20))...... 26 Figure 2: Factors influencing the development of the nasopharyngeal microbiota (figure from van den Broek et al., 2019 (2)) ...... 31 Figure 3: Development of healthy nasopharyngeal microbiota in the first year of life. Figure from Bosch et al, 2017 (49) ...... 32 Figure 4: The pattern of relative abundance difference for each pathobiont (Pettigrew et al, 2012 (3)) ...... 35 Figure 5: Distribution of OTUs in the upper respiratory tract in Australian Indigenous children with OME (Jervis-Bardy et al (1)) ...... 39 Figure 6: Potential mechanisms of bacterial interference ...... 44 Figure 7: Literature search and selection ...... 67 Figure 8: Forest plot showing bacteriology in relation to acute otitis media ...... 78 Figure 9: Forest plot showing bacteriology in relation to otitis media with effusion ...... 80 Figure 10: Forest plot showing bacteriology in relation to chronic suppurative otitis media ...... 82 Figure 11: Recommendations for future research of OM microbiology in indigenous children ...... 87 Figure 12: Participatory Action Research approach ...... 102 Figure 13: Study logo to facilitate community engagement ...... 103 Figure 14: Summary of methods ...... 107 Figure 15: Study workflow ...... 119 Figure 16: Illustration of cumulative species distribution across each of the three anatomical sites ...... 123 Figure 17: Correlation of culture-based species, respiratory viruses and otopathogen loads according “healthy” status. Results filtered for purposes of clarity to show stronger correlations (i.e. absolute Pearson correlation ≥ 0.3 and adjusted p-value ≤ 0.05) ...... 125 Figure 18: Correlation of species detected by culture (excluding otopathogens), respiratory viruses and otopathogen loads according by nose status ...... 127 Figure 19: Principal coordinate analysis for nose status ...... 128 Figure 20: Mean relative microbial abundances of the 20 most abundant genera (or lowest resolved taxonomy level) across all samples illustrating differences between otitis status, community of residence and other key variables. Microbes with lower abundances have been combined in the

‘Other’ (in grey). To improve interpretability, samples have been ordered by Otitis status, Community and Dolosigranulum abundance...... 166 Figure 21: Network correlation analysis showing differences in genera relationships in context of otitis status: A) Never OM; B): HxOM; C) Middle ear effusion; D) TM perforation...... 168 Figure 22: Correlation network analysis of genera in relation to nose health showing differential Dolosigranulum relationships...... 170 Figure 23: Genus-level Principal Component Analysis showing no separation in relation to A) otitis status; B) community of residence; C) season of swab collection; D) number of people residing within the household...... 171 Figure 24: Agar overlay inhibition guide ...... 194 Figure 25: Agar overlay of lactobacillus against Streptococcus pneumoniae, Haemophilus influenzae, and Moraxella catarrhalis ...... 197 Figure 26: Agar overlay of alpha haemolytic streptococcus against Streptococcus pneumoniae, Haemophilus influenzae, and Moraxella catarrhalis ...... 198 Figure 27: Inhibiting and killing effect of alpha streptococcus and lactobacillus on otopathogens. Lactobacillus (dashed lines) are bactericidal against A) H. influenzae, B) and C) M. catarrhalis, and D) S. pneumoniae. Alpha haemolytic streptococcus are bactericidal against S. pneumoniae (D), but not H. influenzae (A), or M. catarrhalis (B and C)...... 199

Supplementary Figures Supplementary Figure 1: Prevalence of respiratory viruses in nasal swabs ...... 155 Supplementary Figure 2: Dolosigranulum significantly correlated with Corynebacterium in Never OM and Moraxella in HxOM...... 185 Supplementary Figure 3: Children with effusion had higher mean relative abundance of Ornithobacterium, compared to never OM...... 186 Supplementary Figure 4: Supplementary testing of alpha streptococcus again H. influenzae and S. pneumoniae ...... 224 Supplementary Figure 5: Cell-free supernatant bacterial interference of pathogens from Indigenous Australian children by commercial probiotic strains LGG and LB21 compared to lactobacillus from the upper airways of Indigenous Australian children ...... 225 Supplementary Figure 6: Bacterial interference of alpha streptococcus against respiratory pathogens using agar overlay; comparison of healthy children compared to those with chronic suppurative otitis media ...... 226

List of Abbreviations used in the thesis AHS Alpha haemolytic streptococci Adj. Adjusted ANI Average nucleotide identity AOM Acute otitis media AOMwP Acute otitis media with perforated tympanic membrane ASV Amplicon sequence variants CASP Critical Appraisal Skills Program CI Confidence interval Clr Centred log-ratio CMV Cytomegalovirus Cq quantification cycle CSOM Chronic suppurative otitis media EAC External auditory canal ENT Ear, nose and throat ET Eustachian tube ERV-3 Endogenous Retrovirus-3 HAdV Human adenovirus HGD Human genomic DNA HMPV Human metapneumovirus HxOM History of OM, but health tympanic membrane at time of collection

LB Lactobacillus MALDI-TOF MS Matrix assisted laser desorption ionization-time of flight mass spectrometry MED Middle ear discharge MEE Middle ear effusion NGS Next generation sequencing NNTB Number needed to treat for beneficial outcome NP Nasopharynx NTHI Non-typeable Haemophilus influenzae OM Otitis media OME Otitis media with effusion OR Odds Ratio

OTU Operational taxonomic unit PA Positive agreement PAR Participatory action research PERMANOVA Permutational multivariate analysis of variance PERMDISP Analysis of multivariate homogeneity PCA Principal Component Analysis PCO Principal Coordinates Analysis PRISMA Preferred Reporting Items for Systematic Reviews and Meta-Analyses qPCR Quantitative PCR rAOM Recurrent acute otitis media RCT Randomised controlled trial rRNA Ribosomal ribonucleic acid RR Risk ratio RSV Respiratory syncytial virus SD Standard deviation TM Tympanic membrane URT Upper Respiratory tract URTI Upper respiratory tract infection WGS Whole genome sequencing WHO World Health Organisation

Chapter 1: Introduction

Otitis Media 1.1.1 Definitions Otitis media (OM) refers to middle ear inflammation and/or infection in, characterised by fluid in the middle ear (4-7). The spectrum of disease is defined in Table 1. Children who have six or more episodes, or three or more episodes in one year are considered “otitis-prone”(8). Table 1: Definitions of otitis media (Kong et al ((7))

Type of OM Definition

Acute otitis media Presence of middle ear fluid with symptoms or signs of (AOM) without suppurative infection, which may include otalgia, fever, perforation irritability, vomiting or diarrhoea.

AOM with Acute suppurative infection with recent discharge from the middle perforation ear or through a tympanostomy tube (within the past 7 days).

Recurrent AOM Recurrent bouts of AOM — three episodes in 6 months or four to (rAOM) five in 12 months.

Otitis media with Presence of middle ear fluid without symptoms or signs of effusion (OME) suppurative infection.

Chronic suppurative A persistent discharge from the middle ear through a tympanic otitis media (CSOM) membrane perforation for more than 6 weeks. CSOM may include a chronic perforation with or without acute or chronic otorrhoea.

1.1.2 Epidemiology and cost of otitis media Aboriginal and Torres Strait Islander (respectively referred to henceforth as Indigenous Australian) children have some of the highest rates of OM in the world (9). Indigenous Australian children experience OM earlier, more frequently and in more severe forms than non-Indigenous Australians (7, 10, 11), with little change in prevalence since the first large ear health surveys in the 1970s (12, 13). During the 2007 Northern Territory Emergency Response, 5473 children were reviewed, only 30% had healthy ears (13). Prevalence peaks in preschool aged children; in remote Northern and Central Australia up to 91% of preschool-aged children had OM in cross-sectional testing (10, 14). Perforation of the TM affected 40% of children in their first 18 months of life and were documented

in children as young as 19 days of age (10). Up to 28% of Indigenous Australian children have CSOM (10, 12, 14-16), well over the 4% breakpoint The World Health Organisation (WHO) considers as a massive public health problem which requires urgent attention (17).

Otitis Media and its sequelae produce substantial economic and societal costs (17). In Australia, the treatment costs of OM in 2008 were between AUS$100 million and AUS$400 million, excluding the costs of complications and comorbidities (18). The estimated net cost of lost wellbeing was between AUS$1.05 billion and AUS$2.6 billion (7). The cost of OM among Indigenous Australians was approximately AUS$38.7 million (18). However, this is likely a gross underestimation of the true cost as the estimates were based on the average for all Australians, whereas the cost of health service provision in remote areas is significantly higher. Furthermore indirect financial costs apply to the child’s family as carer costs and include productivity losses, losses of taxation revenue and travel costs (6).

1.1.3 Pathogenesis of otitis media The pathogenesis of OM is multifactorial and involves the adaptive and native immune systems, Eustachian tube (ET) dysfunction, viral and bacterial load and genetic and environmental factors (19). The development of OM commences with ET dysfunction, impeding normal middle ear aeration and drainage, creating a negative pressure middle ear environment which can lead to transudation of fluid into the middle ear (4). Eustachian tube dysfunction can be caused by mucosal inflammation from an upper respiratory tract infection (URTI), obstruction of the tubal orifice, or other anatomical/ functional pathology (4). OM is often associated with common respiratory tract pathobionts, Streptococcus pneumoniae, non-typeable Haemophilus influenzae (NTHI), and Moraxella catarrhalis (hereby referred to as otopathogens) and respiratory viruses (20, 21) and typically seed from the nasopharynx (NP) (4). OM is a polymicrobial disease and co-infection with >1 otopathogen ± viruses is common, particularly in Indigenous Australian children (22). Transformation into chronic phenotypes, particularly rAOM and CSOM, can occur through several possible avenues including impairment of mucosal immune response, development of biofilms, and/or re-infection of the NP with otopathogens (Figure 1) (4, 20).

Figure 1: Proposed model to explain chronic OM in Indigenous Australian children (Wiertsema & Leach, 2009 (20)).

1.1.4 Risk factors for otitis media For all populations, risk factors for AOM can be divided into host and environmental factors. Host factors include age, premature birth, genetic predisposition, allergies, craniofacial abnormalities, immunodeficiency, and the presence of URTI (4, 23, 24). Environmental risk factors include poor community and domestic infrastructure, overcrowding, exposure to tobacco smoke, and indoor allergens (23, 24). In Indigenous Australian children, early otopathogen colonisation is a well- recognised risk factor (11). There is great diversity between Indigenous Australian communities and consequently OM risk factors vary. However many communities, particular remote communities share common risk factors including, overcrowding, exposure to tobacco smoke and poor domestic infrastructure (23). Breastfeeding is generally considered protective against OM, however a meta- analysis failed to demonstrate that breastfeeding, even for >6 months reduces the risk of developing chronic/ recurrent OM (23, 24). Current data suggests that breastfeeding rates among Indigenous Australians are lower than non-Indigenous Australians (25). A national survey from 2012-13 reported that 83% of Indigenous Australian children aged 0-3 years had been breastfeed, with only 18% continuing to be breastfeed >6 months (25). This data, however, may underrepresent true breastfeeding rates as there may be participation bias.

1.1.5 Current treatment and prevention of otitis media in Indigenous Australian children The treatment of OM differs according to the type of OM. The current treatment guidelines for OM specific to Indigenous Australian children are outlined in Table 2 (5).

Table 2: Treatment guidelines for Indigenous Australian children (5)

Type of OM Treatment

AOM without At least 7 days amoxicillin perforation

AOM with 14 days amoxicillin, continue until discharge resolved. If persistent, perforation change to amoxicillin-clavulanate for further 14-28 days + ear toilet and ciprofloxacin ear drops

Recurrent AOM Amoxicillin for 3-6 months

CSOM Aural toilet and ciprofloxacin ear drops until ear has been dry for > 3 days

Persistent OME Ongoing review and consider referral for grommet surgery (> 3 months)

The current treatment guidelines for Indigenous Australian children are focused on antibiotic therapy, targeting otopathogens. In contrary, most cases of OM in non-Indigenous children are monitored without antibiotics due to poor efficacy, adverse effects (diarrhoea, rashes, abdominal pain), and increasing global awareness of antibiotic resistant microorganisms (26). A Cochrane Review, including both Indigenous and non-Indigenous children, showed that long-term antibiotics can prevent rAOM; however, did not include children with TM perforations or chronic OM (27). Long durations of antibiotics also carry increase risk of antibiotic-associated complications and antibiotic resistance, particularly in patients where treatment compliance may be low. Indigenous Australian children are disproportionately burdened by CSOM. The associated otorrhoea can be effectively treated with topical antibiotics (28); however, the chronic TM perforation remains, providing a conduit for further infection and impacts on hearing. Surgical intervention with tympanoplasty is required to close the TM, but success varies (29). The most effective solution is prevention, both primary prevention and secondary prevention from acute transformation into chronic phenotypes. 27

Currently, vaccination is used to prevent OM. In Australia, pneumococcal vaccination commenced under the National Immunisation Program in mid-2001 using the 7-valent pneumococcal conjugate vaccine (7vPCV) (30). In most Australian states this was replaced by the 13vPCV in 2011(30). Currently 13vPCV is given at 2, 4, 6 months and again at 12-18 months in Indigenous Australian children in high risk areas (30). Children in Australia are also vaccinated again Haemophilus influenzae type b, however OM is most often caused by NTHI (30). Contrary to the success of vaccination programs in other infectious diseases, the impact of vaccinations on OM in Indigenous Australian children has been less effective.

In Australian Indigenous children, vaccination programs targeting S. pneumoniae and H. influenzae have resulted in some reduction in rates of CSOM (PHiD-CV10 vaccination), but these coincided with an increase in OME (14, 31), which still results in hearing loss. A systematic review of 16 RCTs investigating pneumococcal vaccination, including 14,776 Indigenous and non-Indigenous participants from around the world, demonstrated that vaccination reduced the prevalence of S. pneumoniae vaccine serotypes in the NP, but resulted in a concurrent increase in non-vaccine serotypes and had no overall impact on S. pneumoniae or NTHI carriage (32). Many Indigenous Australian infants are colonised with otopathogens before their first vaccination at 2 months of age, which may be contributing to the limited success of current vaccination programs (11). There is an urgent need for more successful prevention of OM and otopathogen colonisation in Indigenous Australian children. To enable the development of successful biomedical preventative treatments, a greater understanding of both the healthy and OM-associated URT microbiota is required.

1.1.6 Complications and consequences of otitis media Globally it is estimated that 21,000 people die each year due to OM-related complications, the majority are <5 years of age (33). Complications from OM broadly fall within three categories; extracranial, intracranial and hearing loss associated morbidity (34). Extracranial complications include mastoiditis, Bezold abscess, facial nerve palsy, sensorineural deafness, tympanic membrane perforation and cholesteatoma (34). Intracranial complications include meningitis, intracranial abscess and sigmoid sinus thrombosis (34). Long-term morbidity from OM arises as a consequence of the associated hearing loss. All phenotypes of OM are associated with a mild to moderate conductive hearing loss, however it is chronic and permanent hearing loss that causes morbidity (34). Otitis media-associated hearing loss can impact significantly on language and social skills 28

development, school attendance and educational outcomes, and has been associated with greater contact with the criminal justice system later in life (7, 35, 36). This is particularly pertinent for Indigenous Australian children who experience OM earlier, more frequently and in more severe forms than non-Indigenous Australians (7, 10, 11).

Upper Respiratory Tract Microbiota 1.2.1 Methods to analyse the upper respiratory tract microbiota Traditionally the URT microbiota was investigated using a limited culture-based technique which falls short in assessment of the microbiome. Using the current standard culture-based techniques, a large portion of microorganisms are ‘unculturable’ resulting in them being undetected (37). ‘Dominant’ species within a culture-based analysis can be misleading, representing organisms which grow most favourable under set conditions, as opposed to those which are most abundant in the environment of interest (37). Culture-based analysis can be less sensitive for detecting bacteria compared to molecular methods (38). The benefits of culture-based analysis sometimes outweigh these limitations. It produces bacterial material for further investigation, which is particularly important in searching for possible probiotic strains. Furthermore, a culturomics approach which uses high-throughput expanded culture with modern identification methods such as Matrix assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) allows for accurate identification to the species-level (39). In response to culture-based limitations, microbiologists started using next-generation sequencing to explore microbial communities, predominantly via bacterial 16S ribosomal ribonucleic acid (rRNA). However, 16S rRNA also has limitations and biases: Samples are prepared using an oligonucleotide primer-based amplification step, which often has insufficient resolution to determine identification down to the species-level; it can fail to detect some phyla; and can skew relative abundance data (40). The upper airways, particularly the middle ear, has a low bacterial biomass resulting in failure of standard PCR and the need for nested PCR, which can further amplify these biases. Pathogenicity is often expressed at the strain-level, therefore microbiological techniques with accurate resolution to only the level of genus has significant short- comings when investigating infectious diseases (2). Specific to the URT, 16S rRNA is particularly poor at differentiating streptococci species, a genus containing both important commensal and pathogenic species in the URT (41) . Metagenomic sequencing does not require PCR and therefore avoids these biases, it includes microorganisms beyond bacteria (e.g. viruses and fungi), and analysis of the microbial genomes beyond the 16S rRNA region means that species-level, and even strain-level identification is possible (42). However, metagenomic sequencing also includes human 29

genomic DNA, and if the ratio of human to microbial DNA is very high, as is the case in the URT, the signal from the microbes can be overwhelmed leading to inaccurate classification (42). No one has yet developed a method to improve the ratio of human to microbial DNA without biasing the sample, and therefore metagenomics sequencing has yet to be reported for the evaluation of the URT microbiome. Furthermore, data from human genomic DNA in Indigenous populations is particularly sensitive, and even if incidentally collected needs to be handled with caution and specific and detailed community consultation and consent is required.

1.2.2 Development of the nasopharyngeal microbiota In the first few years of life, the healthy NP microbiota in non-Indigenous children is distinctly individual and dynamic, influenced by mode of delivery and evolving with age, the season and ingestion of breast milk or formula (43-46) (Figure 2). At birth the infant’s microbiota is largely homogenous across anatomical sites, sparsely populated with bacteria and reflects the mother’s skin or vaginal tract, depending on the mode of delivery (47-49). By day one of life, the NP microbiota becomes dominated by Streptococcus viridans, then rapidly develops niche-differentiation (48). Although a ‘core microbiome’ has not been identified in non-Indigenous children, distinct microbial profiles appear to exist, dominated by key bacteria (44, 48, 49). During the first few years of life children often move between these profiles (Figure 3) (44, 48). On the path to niche-differentiation there is initially a high incidence and abundance of S. aureus, which is then replaced by profiles dominated by either Corynebacterium pseudodiphtheriticum/propinquum, Dolosigranulum pigrum, M. catarrhalis/ nonliquefaciens, S. pneumoniae, and/or H. influenzae (48). The succession and timing of these profiles appear to be critical for NP microbiota stability and propensity for respiratory infections (49). Specifically, NP microbiota dominated early by Corynebacterium/ Dolosigranulum spp. were found to be more stable, related to less respiratory infections in the first year of life, associated with breastfeeding and vaginal delivery, and were depleted in children who had antibiotics (44, 49). Conversely, profiles dominated by Haemophilus spp. or Streptococcus spp. tended to be less stable (44). By three months of age, Moraxella spp. becomes dominant in most children, however children tend to suffer more respiratory tract infections if this ‘maturation’ into a Moraxella spp. profile occurs prematurely and there is less prolonged establishment of Corynebacterium/ Dolosigranulum spp. (49). The current data supports a ‘critical window’ of NP microbiota development that may have the potential to impact respiratory health through childhood, influenced by the child’s environment. Manipulation of this very early NP microbiome may provide a unique opportunity to positively impact a child’s respiratory health, however more data is 30

required. Longitudinal birth cohort studies mapping microbiota development are difficult to conduct and expensive, however cross-sectional studies exploring the NP microbiota in healthy children compared to children with OM and other URTIs provide further insights.

Figure 2: Factors influencing the development of the nasopharyngeal microbiota (figure from van den Broek et al., 2019 (2))

Figure 3: Development of healthy nasopharyngeal microbiota in the first year of life. Figure from Bosch et al, 2017 (49)

1.2.3 Early nasopharyngeal microbiota in Australian Indigenous children There are limited studies exploring the development of the NP microbiota in Indigenous Australian children and no studies exploring the microbiota beyond the otopathogens. However, the data that does exist suggests that most Indigenous Australian children have NP microbiota profiles that may predispose them to respiratory disease and OM early in life. A longitudinal prospective study of 41 Indigenous Australian children from a remote Indigenous community demonstrated that colonisation of the NP with otopathogens occurred at a very early age; M. catarrhalis by 8 days old, NTHI by 10 days and S. pneumoniae by 20 days, median age 38, 50, 58 days, respectively (11). Most children (86%) were colonized with ≥1 otopathogen at their first episode of OM (11). In the 12 non-Indigenous children included, there was no correlation between NP otopathogen colonization and OM, with colonization occurring much later (>200 days for M. catarrhalis and NTHI; none of the non-Indigenous children were colonized with S. pneumoniae) (11). The Indigenous Australian children were more likely to have persistent OM (OM resolution in 5/127 examinations) and otopathogen colonization (clearance of otopathogens in 3/134 examinations), with persistence of OM significantly related to the multiplicity and persistence of otopathogen colonization (11). This study had a small sample size and only used traditional culture-dependent techniques to investigate the NP microbiota, which is not as sensitive as molecular techniques and therefore may underestimate the prevalence of otopathogen colonization. Furthermore, they did not investigate the influence of non-pathogenic bacteria in ear health and disease. However, the results suggest that at least in this Northern Territory community, Indigenous Australian children are developing an unstable NP microbiota, dominated by otopathogens that may lead to early and persistent OM. These data highlight strategies are urgently required to prevent the very early NP colonization of otopathogens in Indigenous Australian children. This early NP colonization with otopathogens may also affect the development of mucosal immunity in Indigenous Australian children.

1.2.4 Development of mucosal immunity The respiratory tract has both physical and immunological defense systems to combat the constant bombardment with external stimulus (50). Resident microbial flora not only contribute to the protection of the host against possible pathogen invasion, they are also integral to the development of a healthy immune system (50, 51). Experiments using germ-free mice have shown the importance of the bacterial microbiota on mediation of immune cell differentiation and subsequent modulation of inflammation, particularly in the gastrointestinal system (51). Through several large 33

cohort studies and meta-analysis it is now well established that birth via caesarian section is related to a higher risk of food allergy, atopy, allergic rhinitis, and asthma (51). The respiratory microbiota, which is heavily influenced by mode of delivery, may be key to this relationship (51). There is data to suggest that disruption of the respiratory microbiome can impact local immunity, infection susceptibility and the development of chronic respiratory inflammatory diseases (52). It is believed that there is a critical period in infancy/childhood when the influence of microbes on the developing immune system establishes a system of homeostasis (51). It would be interesting to investigate the impact of early otopathogen colonization on the developing immune system of Indigenous Australian children and their ability to respond to subsequent colonisation and infection by these otopathogens. To my knowledge, no such studies currently exist.

1.2.5 Factors That Disrupt the Nasopharyngeal Microbiota Otopathogen colonisation In non-Indigenous children, a relationship has been demonstrated between the URT microbiota and the presence of otopathogens, viruses, antibiotics, and pneumococcal vaccination. A cross-sectional study explored the relationship between otopathogens and the normal nasal flora by obtaining nasal swabs from 240 non-Indigenous children age six months to three years who were healthy (n = 73), with URTI alone (n = 95) or with a URTI and AOM (n = 72) (3). Samples were analysed using 16S rRNA gene sequencing (V1-V2), culture for S. pneumoniae, and real-time PCR for otopathogens (3). The study found that the presence of otopathogens was associated with lower levels of diversity and lower abundance of Lactococcus spp. The pattern of relative abundance differed for each otopathogen (Figure 4) (3). Due to the cross-sectional design, this study is unable to clarify whether the reduced diversity resulted in otopathogen colonisation or vice versa. This relationship between otopathogen colonisation and changes in the NP microbiota were supported by a recent prospective longitudinal study of non-Indigenous infants (53). Nasopharyngeal samples were collected from 139 infants monthly for the first six months of life, at nine months old, and during URTI and AOM and analysed using 16S rRNA (V4) gene sequencing (53). Samples that were culture positive for S. pneumoniae, H. influenzae and for ≥2 otopathogens were associated lower bacterial diversity (53). The relationship between the composition of the NP microbiota and otopathogen colonisation were not explored, nor was the temporality of these changes to diversity documented. The reduced NP microbiota diversity in the presence of otopathogen colonisation may reflect otopathogen overgrowth; however, may also represent an existing dysbiosis that facilitates otopathogen

colonisation. Further longitudinal studies are required to investigate the temporality of this relationship and further elucidate changes in the commensal flora.

Figure 4: The pattern of relative abundance difference for each pathobiont (Pettigrew et al, 2012 (3))

Biofilm In vivo and in vitro studies have demonstrated that otopathogens communicate via quorum sensing to promote biofilm production and thus increase their resilience to antimicrobials and host defences. Some strains of H. influenzae can readily form biofilm with other otopathogens and promotes persistence and antibiotics resistance in M. catarrhalis via quorum sensing signal autoinducer-2 (54). In turn, M. catarrhalis strain 7169 can protect H. influenzae against beta-lactam antibiotics, often used in the treatment of OM (54). In the chinchilla model, the presence of NTHI 86-028NP enhanced the biofilm production of S. pneumoniae TIGR4 and induced it towards a translucent phenotype, which was less likely to cause systemic disease, but more adherent to host cell structures (54). This resulted in large and more persistent S. pneumoniae biofilms in the middle ear of the chinchillas, and increased risk of recurrent infections (54). Alloiococcus otitidis (ATCC 51267), a species often detected in middle ear fluid of children with various OM phenotypes, have been shown to form single species and polymicrobial biofilms with H. influenzae, increase resistance to antibiotics in some strains of H. influenzae, and promote H. influenzae survival in depleted 35

conditions (55). Symbiotic relationships within biofilms have been demonstrated in a limited number of otopathogen strains, however these data highlight the potentially complex inter-species relationships that may occur in the pathogenesis of OM and need to be taken into account when developing new preventions and treatments for OM.

Respiratory viruses There has been limited investigation on whether respiratory viruses impact the NP microbiota. A very small (n = 10) study of non-Indigenous adults demonstrated no changes in the NP microbiota following inoculation with rhinovirus, although only seven participants were successfully colonised and three were symptomatic (56). In a study of 96 healthy non-Indigenous children, respiratory viruses were detected in 67% of NP samples with no evidence of changes in NP microbiota (45). For both these studies, identification was achieved to the genus-level only, therefore any changes at the species-level would not have been detected. In contrast, a study of 114 Indigenous Australian children (366 NP samples) used quantitative PCR and found that the combined load of S. pneumoniae, H. influenzae and M. catarrhalis were significantly greater in the presence of any of the respiratory viruses (57). In isolation, only the load of H. influenzae was increased in the presence of respiratory viruses (57). A larger study of 139 non-Indigenous infants in the first year of life also showed increase in the abundance of Moraxella spp. and Streptococcus spp. in the NP during viral colonisation, but only when the child was symptomatic (53). Further studies, using methods to identify the microbiome down to the level of species is required to clarify the impact viral infections have on the NP microbiome.

Vaccination There is evidence that pneumococcal vaccination may induce transient changes in the developing NP microbiota. A randomised controlled trial (RCT) of 97 PCV-7 vaccinated and 103 unvaccinated non-Indigenous infants under two years of age sequenced the V5-V7 16S rRNA region of NP samples and demonstrated that administration of the pneumococcal vaccination induced changes in the entire NP microbiome, not limited to the pneumococcal serotypes (58). Specifically, pneumococcal vaccination resulted in increased abundance of bacteria from genera including Veillonella, Prevotella, Fusobacterium, Leptotrichia, Actinomyces, Rothia, Neisseria, and non- pneumococcal streptococci (58). There were also differences in the cluster composition and distribution between the two groups. Although any differences had dispersed by 24 months of age (58). In non-Indigenous children under two years of age with AOM, administration of PCV-7 was 36

related to reduced prevalence of Corynebacterium and non-pneumococcal Streptococcaceae (59). Contrary to these findings, a small study of Swiss infants (n = 47) found no differences NP diversity between vaccinated and unvaccinated infants at their first and second dose of pneumococcal vaccine (46). This study, however, only analysed the bacteria to the level of family, and as such may have had insufficient resolution to pick up changes in NP microbiota. Furthermore, nasal swabs were collected by the parents and may not be as accurate as samples collected by trained professionals. The impact of vaccination on non-target bacteria suggest that mucosal immunity may be playing a role in shaping the developing NP microbiota. These results need replication in larger cohorts with the consideration of immunological targets to investigate this mechanism further.

Antibiotics There is growing evidence that administration of antibiotics can affect the microbiota widely, not only reducing the otopathogen burden, but also negatively affecting possible ‘protective’ commensal species. In the NP, antibiotic administration has been associated with reductions in health-associated species including Corynebacterium spp., Dolosigranulum spp., and Streptococcaceae (3, 49, 59). Without the defense of a healthy microbial flora the host becomes susceptible to re-infection. A study using culture-based analysis of middle ear samples obtained from tympanocentesis in non-Indigenous children found that most early recurrences of bacterial AOM following treatment with antibiotics was associated with a different otopathogen (60). Although an interesting finding, this result needs to be replicated to further clarify the relationship between antibiotic use and early relapse of bacterial AOM. However, it does suggest that preservation or replenishment of health-associated bacteria may prevent recurrence and requires further investigation.

Upper Respiratory and Middle Ear Microbiota in Relation to Otitis Media 1.3.1 Bacteriology of otitis media Otitis media is a polypathogenic disease commonly associated with three main otopathogens; S. pneumoniae, M. catarrhalis, H. influenzae (61). A systematic review analysed the results of 66 culture-based studies exploring middle ear fluid of non-Indigenous children with OM (62). In AOM, S. pneumoniae was significantly more prevalent than H. influenzae (pooled analysis p < 0.036); with an average of 27.8% of patients positive for S. pneumoniae compared to 23.1% for H. influenzae (62). In rAOM/ AOM not responding to treatment, H. influenzae was more common, with an average of 22.8% positive cases, compared to 18.8% for S. pneumoniae (62). Middle ear 37

samples from OME were less likely to grow otopathogens, with H. influenzae recovered in 11.6% and S. pneumoniae in 6.5% (62). M. catarrhalis was less frequent in all types of OM (62). Although this systematic review provided valuable data regarding the bacteriology of OM globally, Indigenous Australian populations were not included.

In Indigenous Australian children, studies predominantly used culture-dependent methods to investigate bacteriology in relation to OM. Acute OM in Indigenous Australian children was associated with polymicrobial otopathogen colonisation in the NP, while H. influenzae was most prevalent in middle ear effusion (MEE) (57, 63, 64). Otitis media with effusion was related to high rates of S. pneumoniae and H. influenzae detection in the NP and M. catarrhalis in MEE, although was rarely associated with polymicrobial NP colonisation, (1, 15, 57, 65-67). These studies were limited by the absence of healthy controls, overlapping populations and narrow investigation of the microbiota (i.e. few studies explored beyond the 3 main otopathogens). Furthermore, there is great diversity across Indigenous populations in Australian. The majority of the microbial studies in Indigenous Australian children were from a confined region of Australia. We do not yet know whether these results can be safely generalised to other Indigenous Australian children, who are often from vastly different environments. More microbiological research is required, including Indigenous Australian children from across the diversity of Australian communities and ideally using advanced microbiological techniques capable exploring the entire microbiota.

Several recent studies have used 16S rRNA gene sequencing to explore the microbiota of the MEE and adenoids/NP in children with OME, AOM, rAOM and CSOM (1, 68-72). One of these studies focused on 11 Indigenous Australian children with OME (1). The results confirmed the dominant role of the main otopathogens in all OM phenotypes across both Indigenous and non-Indigenous cohorts; however, beyond the otopathogens, there was significant variability in the middle ear microbiota of children with OM (1, 69-71). These culture-independent methods detected other bacteria not traditionally associated with OM, including from the genera Alloiococcus, Staphylococcus, Pseudomonas, and Turicella in OME (1, 69, 73-75), Alloiococcus in CSOM (71), and Alloiococcus, Turicella, and Staphylococcus in AOM and rAOM (70, 72), the significance of which is equivocal as these bacteria have also been identified in the healthy external auditory canal (EAC) (73, 76). It is important to note that across these studies populations varied greatly in terms of geography, age, social disadvantage and associated population-wide OM prevalence. The studies that compared NP/ adenoids to MEE showed significantly different microbiota between these sites

(Figure 5) (1, 69, 72, 74). This suggests that there may be reservoirs of potential pathogens, other than the NP, that are playing a greater role than the NP/ adenoids in some children. Lappan et al took both MEE and EAC swabs in children with rAOM and showed greater concordance between MEE and EAC than MEE and NP samples (72). Traditionally the EAC has been viewed as a potential source of secondary infection when the TM is perforated, but not in an intact TM. The Lappan et al study did not stipulate whether the participants previously had TM perforations. Data is still emerging and more research using next generation sequencing (NGS) approaches across anatomical sites for the different OM phenotypes is required.

Figure 5: Distribution of OTUs in the upper respiratory tract in Australian Indigenous children with OME (Jervis-Bardy et al (1))

Studies focusing on the differences in the NP microbiota of healthy children compared to those with AOM show contrasting results (3, 53, 59, 69, 77). A prospective longitudinal study of 139 infants in

the first year of life, (n = 971 NP samples), demonstrated infants with AOM had higher average abundance of bacteria from genera Haemophilus, Enterobacter, and Yersinia and lower average abundance of Corynebacterium and Pseudomonas compared to healthy infants (53). The NP microbiota of children with AOM was less diverse than that of healthy children (3, 59). Other studies have shown that compared to children with AOM, the NP microbiota of healthy children has higher average abundance of bacteria from the families Staphylococcaceae, Flavobacteriaceae, Carnobacteriaceae, Comamonadaceae (59) and genera Lactococcus, Propionibacterium, Corynebacterium and Dolosigranulum (3). While some studies showed no difference in the average abundance of bacteria from the families Moraxellaceae and Streptococcaceae (59) or genera Moraxella and Haemophilus (69, 77) between healthy children and children with AOM. Similarly, Chan and colleagues found no differences in the NP microbiota between healthy controls (n = 10) and children with OME (n = 23) (69). No studies were able to analyse the microbiota down to the species-level, which may be an important limitation. NP colonisation with otopathogens is a risk for OM, however the influence of the remainder of the microbiota is equivocal. Overall, there is a suggestion that, at least in AOM, there are differences in the diversity and composition of the NP microbiota when compared to healthy peers; investigation of these differences in the middle ear space poses a greater challenge.

Only one study has attempted to compare the middle ear microbiota in CSOM to the middle ear of healthy controls (71). They found that the healthy middle ear was composed of bacteria from the genera Novosphingobium, Staphylococcus, Streptococcus, Escherichia-Shigella, and Burkholderia, which was consistent between patients, while CSOM middle ear discharge was more diverse and dominated by Haemophilus, Staphylococcus, Alloiococcus, and Streptococcus. The data from the healthy middle ear need to be interpreted with caution. This study did not consider the bacterial biomass and therefore the data may include contaminant reads; the species found in the middle ear in the healthy controls have been previously reported as common contaminants in the 16S rRNA pipeline (78). The majority of normal middle samples came from adults and thus the data may not be generalisable to children. Further studies are required to corroborate the outcomes of this study.

The culture-based analysis and NGS data suggests that the pathogenesis of OM is likely due to a dysbiosis of the URT; an imbalance between the protective commensal flora and the otopathogens. To date, the literature is informative in highlighting that differences potentially exist between the NP microbiome of children with AOM and healthy children; however, it is less clinically relevant

as it fails to identify the bacteria to the species-level. Further research is needed, ideally using techniques to analyse the microbiota down to the level of species, to delineate the differences in NP microbiota in relation to health and OM. Most of these studies neglected to consider the contribution of respiratory viruses and biofilm, not only as a confounding variable for the pathogenesis of OM, but also their influence on the microbiota.

1.3.2 Virology of otitis media Respiratory viruses are important contributors to OM pathogenesis and include respiratory syncytial virus (RSV), parainfluenza, rhinovirus, influenza, enterovirus and adenovirus (HAdV), often in combination with otopathogens (61, 79). Two studies focusing on Indigenous Australian children examined viruses in relation to AOM and OME (11, 57). In 175 Indigenous Australian children with OME, 58% of NP samples were positive for respiratory viruses, predominantly rhinovirus (39%) followed by polyomavirus (11%) (57). In children with AOM, 68% of the NP swabs were positive for any respiratory virus, mostly rhinovirus (41%), followed by HAdV (19%), and polyomavirus (17%) (57). Only HAdV was significantly related to ear state (57). In another small longitudinal study of Indigenous Australian infants, 9/41 were positive for viruses, including rhinovirus (5/41), HAdV type 6 (4/41), influenza A & parainfluenza type 1 and 3 (1/41 each), and cytomegalovirus (24/41) (11). All were collected after onset of OM and after bacterial colonization (11). Unfortunately, neither of these studies tested for viruses in the MEE. In a recent Australian study of 93 non-Indigenous children with rAOM and 103 healthy control cases, RSV, HAdV and human metapneumovirus (HMPV) were detected more frequently in the NP of children with rAOM than controls (72). Concordance between viral detection in the NP and in the MEE of children with rAOM was low, rhinovirus had the highest concordance of 44.4% (72). One study in non- Indigenous children assessed for viruses in the MEE of 116 samples from children with recurrent AOM/ OME and found 28% were positive for viruses only, and 27% contained both viruses and bacteria (80). Interestingly, the presence of viruses in MME was not associated with increased inflammatory response in the middle ear (80). This suggest that detection of viruses in the MEE may reflect a transient migration from the NP, rather than directly infecting the middle ear space. There are several other mechanisms in which viruses can contribute to the pathogenesis of OM including inflammation of the mucosa leading the ET dysfunction, or as outlined above, disrupting the NP microbiota, and/or by facilitating otopathogen overgrowth in the NP and subsequent seeding to the NP (81). Furthermore, the presence of viruses in the MEE, particularly rhinovirus may facilitate antibiotic failure in children co-infected with bacteria (82). 41

1.3.3 Biofilm and intracellular bacteria in otitis media Biofilms are heterogeneous clusters of organised microbial cells enclosed in a matrix of polysaccharide material (83). Strains of the three main otopathogens readily form biofilms in vitro and in vivo (84) and have been found on the middle ear mucosa of non-Indigenous children with rAOM and OME (84-86) and Greenlandic adults and children with CSOM and OME (86, 87), while they are absent in the middle ear of healthy control children (85, 88). Although these biofilms were primarily composed of otopathogens; Pseudomonas aeruginosa and S. aureus were also reported, and many of the biofilms were polymicrobial (85, 88, 89). Fungal elements have been found within biofilms from patients with chronic sinusitis (90), however have not been reported in OM. The adenoids of non-Indigenous patients with rAOM (n = 10) and OME (n = 10) are covered with bacterial biofilms, with almost the entire adenoid of children with rAOM (mean coverage = 98%) covered in biofilm compared to 30% coverage in children with OME and 0.1% in healthy controls (n = 10) (91). The adenoids of children with OME/rAOM (n = 45) are more likely to be colonised with biofilm-producing otopathogen strains compared to healthy children, particularly at the ET meatus (92). Bacteria within biofilms are more resistant to both host defences and antibiotic treatment and thought to contribute significantly to chronicity and treatment failure in chronic forms of OM (85).

In contradiction to this theory, Jensen and colleagues treated 21 Native Greenlandic children with CSOM and middle biofilms with 7-14 days of aural saline irrigation and ciprofloxacin ear drops (93). After initial treatment, 90% of the patients had dry ears. Thirteen patients had recurrence, eight of which had new multi-species infections on examination of MEE (93), challenging the prevailing theory that middle ear biofilm is a major contributor to treatment failure in chronic forms of OM. This finding suggests that planktonically shed bacteria from the NP ± EAC are responsible for recurrence of CSOM in this population. Furthermore, this outcome may not be generalisable to children with rAOM/ OME as the pathophysiology is altered with a perforated TM. The results of this study require verification in a larger population.

Intracellular bacteria has been demonstrated in children with OME and recurrent AOM (85, 94). In a study of 11 middle ear biopsies from 17 children with OME, 36% were found to have gram- positive cocci, likely S. pneumoniae, within the middle ear epithelial cells, despite an absence of biofilm and only one culture positive for H. influenzae (94). The cocci where predominantly in the 42

vacuoles of mucus-secreting cells with evidence of expulsion into the middle ear fluid during mucous exocytosis (94). In another study of 20 children with OME and/or rAOM, intracellular bacteria were found in 12 middle ear biopsies, nine of which had overlying biofilm (85). They found all three otopathogens intracellularly with evidence of bacteria in the epithelial vacuoles (85). Intracellular organisms readily escape host defences and antimicrobials, creating a conduit for re- infection of the middle ear space following eradication of middle ear biofilm or planktonic bacteria. As mentioned above, there is often a high degree of concordance between EAC and MEE when the TM is intact. It may be possible that bacteria can move transcellularly from the EAC into the middle ear space. This possibility has yet to be explored.

Bacterial Interference in the Upper Respiratory Tract Bacterial interference is defined as “a dynamic antagonistic interaction between at least two organisms that affects the life cycle of each”(95). Microorganisms on the mucosal surface of the URT interact in a multitude of ways, one of which is antagonistically, competing for ecological space and interfering with each other’s growth (96). It is believed that this ‘bacterial interference’ by normal mucosal flora can prevent colonisation and proliferation with potential respiratory pathogens, thereby maintaining respiratory health (96). Bacterial interference as a means of preventing disease has been reported since the late 1800s (97), however has struggled to gain traction and translation into clinical practice. Studies in non-Indigenous populations have shown significantly more commensal bacterial with interfering properties against otopathogens in the NP of children not prone to OM when compared the children prone to OM (96). These include aerobic alpha haemolytic streptococci (AHS) (mostly Streptococcus mitis and Streptococcus sanguinis) and anaerobic streptococci (Peptostreptococcus anaerobius) and Prevotella melaninogenica (96). As early as the 1970s Viridians Streptococci (part of the AHS group) were shown to inhibit a range of potential pathogens in vitro including Neisseria meningitides, Moraxella spp., Beta-haemolytic Streptococci, Corynebacterium diphtheriae, S. pneumoniae, S. aureus, and Escherichia coli (96, 98). In vitro studies have confirmed significant bactericidal effects from various strains of AHS against S. pneumoniae, NTHI, and M. catarrhalis, which are strain specific and dependent on the method used to test for bacterial interference (95). In general, bacterial interference is achieved by mechanisms such as occupying specific sites on the epithelial cell surfaces, thus preventing the adherence of pathogens, changes in the microenvironment by, for example, lowering of pH, production of antagonistic substances, and competition for nutritional substances (Figure 6) (96). Alpha haemolytic streptococci owe their inhibitory potential to their production of bacteriocins and 43

other inhibitory substances including ‘viridins’, which have inhibitory activity against gram- positive and gram-negative bacteria (96). Corynebacterium spp. is URT commensal frequently found in the URT in children who aren’t colonized by S. pneumoniae (99). In vitro, Corynebacterium accolens was shown to inhibit the growth of S. pneumoniae by hydrolysing skin surface triacylglycerols to produce anti-pneumococcal free fatty acids (99). This is an example how a commensal microbe uses human resources to positively shape the URT microbiota. Local administration of probiotics is the application of bacterial interference to treat and prevent infections in vivo.

Figure 6: Potential mechanisms of bacterial interference

Probiotics: Successes in Medicine Probiotics are “live microorganisms which, when administered in adequate amounts, confer a health benefit on the host”(100). In 1958, a landmark paper described the use of faecal transplantation to treat patients with Clostridium difficile enterocolitis, restoring the dysbiosis caused by antibiotics, and resulting in dramatic health improvement (101). Despite this success, no further research was conducted on faecal transplantation for another 50 years. Recently, faecal transplantation was tested in a RCT for the treatment of C. difficile enterocolitis against vancomycin and vancomycin plus bowel lavage (102). This study was stopped after the interim analysis. Of the 16 patients who received the faecal transplantation, 13 had resolution of symptoms after the first infusion, another two after second transfusion from a different donor. This compared to resolution of symptoms in 44

just four of the 13 patients receiving vancomycin and three of the 13 receiving vancomycin and bowel lavage (102). This result has been replicated in a number of RCTs (103), and is now offered as treatment for appropriate patients. This is a striking example of where the use of probiotics has been more effective than antibiotics.

Probiotics have also demonstrated impressive results when used in premature infants. Prophylactic administration of probiotics containing Lactobacillus spp. alone or in combination with Bifidobacterium spp. via enteral feeding has been shown to significantly reduced the incidence of severe necrotizing enterocolitis and all causes of mortality in preterm infants (104). There were no reports of systemic infection with probiotic species (104). Probiotic prophylaxis is now routinely used in many neonatal intensive care units around the world.

Beyond the neonatal unit, the administration of probiotics has been shown to reduce death, sepsis and lower respiratory tract infections in disadvantaged infants in India. In a landmark randomised double-blinded placebo-controlled clinical trial investigators gave a ‘synbiotic’ containing Lactobacillus plantarum and fructooligosaccharide to 4556 Indian infants within the first week of life for 7 days and followed the infants for the first 60 days of life (105). They found a significant reduction in death or sepsis in the synbiotic group (n = 123 (5.4%)), compared to placebo (n = 206 (9.0%), p < 0.001). Infants who were given the synbiotic had fewer lower respiratory tract infections (n = 92 (4.0%)) compared to placebo (n = 139 (6.1%), p = 0.002) (105).

Probiotics in the Upper Respiratory Tract Roos et al provided the proof of concept for probiotics in the URT, demonstrating that commensal flora from URT of healthy individuals could prevent recurrence of streptococcal tonsillitis (106). The authors isolated four strains of AHS from the throats of healthy individuals that had a strong ability to inhibit the growth of Group A Streptococcus in vitro. One hundred and thirty children with recurrent streptococcal tonsillitis were treated with antibiotics and then randomized to receive a daily throat spray containing either the probiotic strains or a placebo for 10 days (107). In the eight weeks post baseline, bacteriologically-verified tonsillitis recurrence occurred in only 2% (1/50) of the children in the probiotic treatment group, whilst it occurred in a much larger proportion, 23% (14/61), of children in the placebo group (107). Having demonstrated such efficacy in recurrent streptococcal tonsillitis, the authors applied the concept to OM.

1.6.1 Local administration of probiotics in OM There have been several RCTs investigating the local administration of probiotics to treat/ prevent OM in otitis-prone children (Table 3). Roos et al appear to be the first to investigate the use of a probiotic treatment made of bacterial strains with a demonstrated ability to inhibit the growth of otopathogens (8). Roos and colleagues isolated AHS from the opening of the ET of healthy children and identified five strains that were powerful inhibitors of S. pneumoniae, M. catarrhalis, and S. pyogenes (8). They developed a probiotic nasal spray containing two strains of Streptococcus sanguinis, two strains of Streptococcus mitis and one strain of Streptococcus oralis. One-hundred and eight otitis-prone children were randomized to receive either the probiotic or a placebo. All children received 10 days of antibiotics, followed by either the probiotic or placebo nasal spray twice daily for 10 days, then for another 10 days after 55-60 days. There was a significant reduction in the rate of recurrence amongst those who received the probiotic, with 42% (22/53) children in the treatment group having healthy ears, compared to 22% (12/55) in the placebo group (8). Significantly less children in the treatment group had OME at the last visit (31%; 10/32), compared to those in the placebo group (56%; 15/27) (8).

An attempt to replicate the Roos result by Tano et al, whom did not pre-treat participants with antibiotics, was unsuccessful (108). Specifically, Tano et al conducted a smaller RCT (n = 36) testing a probiotic treatment comprised of a mixture of AHS strains that had demonstrated in vitro a good ability to inhibit the growth of otopathogens as well as good adherence to adenoid epithelial cells (108). The study recruited children with a history of rAOM. Despite four months of treatment, there was no difference in the number of children with AOM, 44% (7/16) in the treatment groups vs. 40% (8/20) in the placebo group (108). Prior dosing with antibiotics may support probiotic colonisation through reducing bacterial load, providing ecological space and reduced competition for binding sites, and through their anti-inflammatory properties (particularly macrolide 14- and 15- members), reducing host defence against probiotic colonisation (109). Testing the NP for the presence and abundance of the probiotic strains after treatment would have informed on whether this contributed to the failure of the treatment.

A noteworthy RCT conducted by Marchisio et al (110) provides insight into the ability of a probiotic treatment to colonize the participant. Otitis-prone children were randomized to test a probiotic treatment consisting of Streptococcus salivarius 24SMB; a strain obtained from the nose of a healthy child and shown to produce bacteriocin-like substances with activity against

otopathogens (111). The children were treated with antibiotics and then randomized to receive either the probiotic or placebo nasal spray twice daily for five consecutive days each month for three consecutive months. Quantitative PCR of NP swab samples was used to detect the presence of S. salivarius 24SMB. After treatment, the proportion of children who did not experience AOM in the probiotic group (30%; 15/50) was double what it was in the control group (15%; 7/47) (p = 0.075) (110). When the authors limited their analysis to children in the treatment group who yielded an NP swab sample that contained the probiotic, they found that children who were successfully colonized with S. salivarius 24SMB were significantly more likely to have no episodes of AOM (43%; 12/28); compared to those who were not colonized with S. salivarius 24SMB (14%; 3/22) (p = 0.03) (110). These data suggest that colonization with the probiotic strain may be playing a role in treatment efficacy.

The above studies focused on the prevention of AOM, one study by Skovbjerg et al, explored the use of probiotics to treat OME (112). A double-blinded pilot study was conducted on 60 children with OME, using a nasal spray containing either S. sanguinis, Lactobacillus rhamnosus or placebo for 10 days; no antibiotics were given prior to the commencement of the treatment (112). There were significantly more children with complete or significant clinical recovery following treatment with S. sanguinis (7/19) compared to placebo (1/17) (p < 0.05), but not for those who received L. rhamnosus (3/18) compared to placebo (p = 0.60) (112). Interestingly, there was no evidence of colonization with S. sanguinis in the NP following treatment, although culture-based analysis was used, which may not have sufficient sensitivity. Alternatively, the probiotic may have been exerting its action via the immune system. The possible importance of this immunological mediation was recently demonstrated in a murine model of OM investigating the ability of the rodent commensal Muribacter muris (from the same family of NTHi) to prevent colonisation and infection with NTHi (113). In 12 mice pre-treated with M. muris, only one developed OM after inoculation with NTHi at day five, compared with 8/15 with no pre-treatment (113). At day five, the pre-treated mice had lower NTHi loads, lower inflammatory markers, clinical score and weight loss, despite the absence of M. muris nasal colonisation, indicating these effects were mediated by immune modulation (113). Furthermore, inoculation with M. muris stimulated an initial increase in interleukin-6 and keratinocyte chemoattractant on day 0, without an increased in commensal density, suggesting activation of innate immunity (113). These mechanisms have not yet been examined in humans.

Table 3: Randomized controlled trials investigating local (intranasal) administration of probiotics to prevent/ treat otitis media

Study n Age Population Treatment Primary Outcome Results p-value Measure(s) Roos 2001 (8) 108 6 months – Otitis-prone Local: Streptococcus mitis, S. No AOM + Treatment (n=53) = 42% 0.02 6 years sanguinis, Streptococci oralis normal TM Placebo (n=55) = 22% OME Treatment = 31% 0.05 Placebo = 56% Tano 2002 (108) 36 <4 years Otitis-prone Local: S. mitis, S. sanguinis, S. Recurrence of Treatment (n=16) = 44% n.s oralis AOM Placebo (n=20) = 40% Skovbjerg 2008 54 1-8 years Otitis-prone Local: S. sanguinis OR L. Resolution OME S. sanguinis (n=19) = 37% 0.05 (112) rhamnosus L. rhamnosus (n=18) = 17% 0.06 Placebo (n=17) = 6% Marchisio 2015 97 1-5 years Otitis-prone Local: Streptococcus salivarius Recurrence of Treatment (n=50) = 70% 0.08 (110) 24SMB AOM Placebo (n=47) = 85% NP colonized (n=28)= 57% 0.03 not colonized (n=22) = 86%

Note: AOM, acute otitis media; NP, nasopharynx; n.s., non-significant (no p-value reported); TM, tympanic membrane.

1.6.2 Systemic administration of probiotics in OM These studies have explored local probiotic administration. Other studies have investigated the systemic probiotic administration on the prevention of OM (Table 4). In healthy children, several RCTs used L. rhamnosus GG and/or Bifidobacterium lactis BB-12 orally to investigate whether OM could be prevented. Milk supplemented with L. rhamnosus GG was given to 513 day care children for seven months and showed no significant differences in the number of episodes of AOM in the probiotic (31%; 79/252) compared to placebo groups (39%; 101/261) (p = 0.08) (114). Stecksen-Blicks et al gave 186 healthy children milk with L. rhamnosus GG for 21 months and measured the number of days with OM (115). They found children who received the probiotic milk (n = 110) had fewer days with OM (0.5) compared to placebo (n = 76) (1.0) (p = 0.003) (115). In a younger cohort of infants <2 months old, L. rhamnosus GG + B. lactis BB-12 supplemented infant formula was given for 12 months and shown to significantly reduce the incidence of AOM in the first 7 months of life, 22% (7/32) had AOM in treatment group vs. 50% (20/40) in the placebo group (p = 0.014)) (116). However, there were no statistically significant differences in the number of children who suffered rAOM (≥3 episodes); four (13%) in the treatment group and 10 (25%) in the placebo group (p = 0.183) (116). Another study gave B. lactis BB-12 to infants via slow-release tablets for seven months and showed no significant difference in the incidence of self-reported OM compared to placebo; (26%; 9/34 and 17%; 6/35 respectively (p = 0.455)) (117). Whether L. rhamnosus GG and B. lactis BB-12 can prevent OM in healthy children is equivocal. L. rhamnosus GG was originally isolated from the healthy gastrointestinal system (118) and B. lactis BB-12 was a dairy culture (119). Therefore, their poor efficacy in preventing disease from respiratory pathogens may be due to them being used beyond their niche environment.

One study by Di Pierro et al investigated whether oral administration of a probiotic strain niche to the URT, S. salivarius K12, could prevent AOM in healthy children attending kindergarten (120). They delivered the probiotics via dissolving oral tablet and found a significant reduction in the incidence of AOM in the treatment (44%; 49/111) compared to placebo group (80%; 89/111) and the number of episodes of AOM, 53 in the treatment group vs 101 in the placebo group (120). S. salivarius has been shown to adhere to the URT and strongly inhibit the growth of the main respiratory pathogens in vitro. This study suggests that there may be a role for niche specific probiotic strains in preventing AOM in healthy children and further research is required.

In otitis-prone children a range of probiotic mixtures have been administered systemically to determine whether recurrence of OM can be prevented. Two hundred and sixty-nine otitis-prone children were given a probiotic capsule containing two strains of L. rhamnosus GG, Bifidobacterium breve 99, and Propionibacterium freudenreichii spp. shermani JS (121). There were no differences in the incidence of AOM; 72% (n = 135) vs. 65% (n = 134) for treatment vs. placebo (OR = 1.48 (95% CI 0.87 - 2.52)) or rAOM (18% vs 17% respectively; OR = 1.04 (95% CI 0.55 - 1.96)) (121). Correspondingly, 166 “high-risk” children were given formula supplemented with Streptococcus thermophilus, S. salivarius, L. rhamnosus LPR and a prebiotic for 12 months and found no difference in the number of episodes of AOM in the treatment (249) (n = 83) compared to control (237) (n = 83) (122). Cumulatively, these results demonstrate that ingestion of non-specific systemic probiotics does not prevent AOM in healthy children or reduce the number of episodes in otitis-prone/ “high-risk” children, regardless of the duration of treatment. In contrast, local administration using strains isolated from the URT of healthy children showed more promising results in preventing OM in otitis-prone children.

In review of the evidence for the use of probiotics in AOM, a meta-analysis was conducted of 16 RCTs, 11 using Lactobacillus-based probiotic, and six using Streptococcus-based probiotic (123). The review combined all probiotic trials and showed that less children who were given a probiotic experienced ³1 of AOM (risk ratio (RR) 0.7 (95% CI 0.63 - 0.93, number needed to treat for beneficial outcome (NNTB) = 10) and this effect was stronger in non-otitis-prone children compared to otitis-prone children (123). Children who received the probiotic treatment also had less antibiotics for any infection compared to children who received the placebo (RR 0.66 (95% CI 0.51 - 0.86 NNTB = 8) (123). There were no differences in the children who reported adverse events between the probiotic and placebo groups (4 trials). There are potential issues with combining these RCTs that use difference probiotic species and strains, different doses and different administration regimens; however, it does highlight that probiotics are a safe and potentially effective treatment for prevention of AOM in children.

The use of probiotics to prevent and treat OM shows promise and proof of concept exists. For such a probiotic to be effective it likely needs to be niche-specific, demonstrate in vitro bacterial inference against the main otopathogens, and able to colonize the NP. Although, there may also be a systemic effect through activation of the immune system. In otitis-prone children, antibiotics may be required prior to giving the probiotic to reduce the overall bacterial load and facilitate 51

colonization. Further research is required in both the prevention and treatment of all types of OM, and particularly in Indigenous communities. To facilitate such research it important to investigate the complete microbiota in the URT, beyond the otopathogens which have been the historical focus. Infection is a sign of dysbiosis and pathogens in isolation can no longer be considered solely responsible.

Table 4: Randomized controlled trials investigating systemic (oral) administration of probiotics to prevent/ treat otitis media

Study n Age Population Treatment Primary Outcome Result(s) p- Measure(s) value Hatakka 2001 513 1-6 years Healthy Systemic: Lactobacillus Incidence AOM Treatment (n=252) = 31% 0.08 (114) rhamnosus GG Placebo (n=261) = 39% Hatakka 2007 269 10 mo-6 Otitis- Systemic: L. rhamnosus GG and Incidence AOM Treatment (n=135) = 72 % n.s (121) years prone LC705, Bifidobacterium breve Placebo (n=134) = 65% 99 and Propionibacterium freudenreichii JS Rautava 2008 72 <2 Healthy Systemic: L. rhamnosus GG & Incidence AOM 1st 7 Treatment (n=32) = 22% 0.01 (116) months Bifidobacterium lactis BB-12 months Placebo (n=40) = 50% Recurrent (>3) AOM in Treatment = 13% 0.18 1st year Placebo = 25% Stecksen-Blicks 186 1-5 years Healthy Systemic: L. rhamnosus LB21, No. days with AOM Treatment (n=110) = 0.5 <0.01 2009 (115) fluoride Placebo (n=76) = 1.0 Taipale 2011 69 1-2 Healthy Systemic: B. lactis BB-12 Incidence AOM Treatment (n=34) = 26% 0.50 (117) months Placebo (n=35) = 17% Cohen 2013 (122) 166 7-13 Healthy Systemic: Streptococcus No. episodes of AOM Treatment (n=83) = 249 0.80 months thermophiles, S. salivarius, L. Placebo (n=83) = 237

rhamnosus + preB Incidence recurrent AOM Treatment = 30% 0.90 Raftilose/Raftiline Placebo = 30% Di Peirro 2016 222 3 years Healthy Systemic: S. salivarius K12 Incidence AOM Treatment (n=111) = 44% <0.01 (120) (BLIS K12) Placebo (n=111) = 80% Note: AOM, acute otitis media; n.s., non-significant (no p-value reported).

Non-probiotic commensals to prevent otitis media Data from microbiota research and bacterial interference are merging to identify novel commensals for use as therapeutics to prevent a variety of infectious diseases. Examples are listed in Table 5, adapted from Marsh et al (41). These novel commensal bacteria are being investigated for their potential to be directly inoculated live into the URT, but also for bacteriocin-like substances that can be utilized as antimicrobials. This broadens the possibilities for prevention of OM beyond the current antibiotic and vaccination strategies.

Table 5: Novel commensal species in development to prevent infectious diseases, from Marsh et al (41)

Disease Target pathogen Commensal Species Development Ref stage Otitis media S. pneumoniae AHS Phase II - (110) children Otitis media H. influenzae Haemophilus Pre-clinical (124) haemolyticus Pharyngitis S. pyogenes Streptococcus salivarius Licensed (125) product Pneumococcal S. pneumoniae Corynebacterium Pre-clinical (99) infections accolens Pneumonia S. pneumoniae Streptococcus mitis Pre-clinical (126) Meningitis Neisseria meningitidis Neisseria lactamica Pre-clinical (127, 128) Bacteraemia S. aureus Staphylococcus Phase I - adults (129) lugdunensis Salmonellosis Salmonella Escherichia coli Pre-clinical (130) typhi(murium)

Conclusion There is a significant, ongoing burden of disease caused by OM in the Indigenous Australian population, despite vaccinations and antibiotic treatment. The causes are multi-factorial and include social, environmental, biomedical and possible immune and genetic factors (20). A multi-pronged

approach to the treatment and prevention of OM is required; from a biomedical perspective “strategies to reduce the exposure to high doses of multiple OM causing bacteria are urgently needed” (20). Antibiotics may temporarily resolve OM and reduce the number of otopathogens; however they likely disrupt the normal protective flora and allow for recolonization with otopathogens and recurrent infections, in addition to growing global concerns regarding antimicrobial resistance. An alternative, more effective treatment is required. Local administration of a probiotic that is niche-specific for the URT and demonstrates the ability to interfere with the growth of otopathogens needs to be evaluated as an alternative treatment for the refractory problem of OM, particularly in developing nations and Indigenous populations. However, for such a probiotic to be developed further research is required to explore the URT microbiota in Indigenous children in health and with OM.

Aims and Hypotheses As demonstrated above, Indigenous Australian children continue to suffer a high burden of disease from OM, impacting on hearing, child development, educational and employment outcomes. The disease prevalence and severity are essentially unchanged despite vaccination and antibiotic treatments. The is an urgent need for innovative treatments, aimed at preventing OM in Indigenous Australian children, either through primary preventing or secondary prevention of recurrence/ chronic transformation. Research in non-Indigenous children suggests that probiotics may be a safe and effective treatment for the prevention of OM, however no one has yet to explore this in Indigenous children. Effective probiotic strains used in the URT thus far have been obtained from healthy European children and used to treat/ prevent AOM/rAOM and OME, the dominant phenotypes in affluent communities. Indigenous Australian children however are vastly different from European children in terms of geography, environmental influences, population-wide OM prevalence, and dominance of CSOM phenotype. Theoretically probiotic strains obtained from the URT of healthy Indigenous Australian children, who do not get OM despite endemic levels of CSOM within the community, are going to be more effective at preventing OM than strains obtained from children in Europe. The ultimate aim of this thesis therefore is to identify protective bacterial strains than can be developed into a probiotic specifically to prevent OM in Indigenous Australian children. The hypothesis is as follows:

Healthy Indigenous Australian children will have bacterial strains in their URT that have the ability to inhibit the growth of otopathogens in vitro and are less prevalent/ less abundant in the URT of Indigenous Australian children with OM/ URTIs.

This hypothesis will be explored through the following aims:

1.9.1 Aim 1: To explore the current microbial data in relation to otitis media in indigenous populations around the world The initial aim of this thesis is to canvas the microbial data relating to OM in indigenous populations globally to establish the current knowledge and identify potential gaps in the literature. This will be achieved by conducting a systematic review with a focus on the URT microbiota in relation to OM.

Hypothesis: There will be strong evidence supporting the association between the three main otopathogens and OM in indigenous populations, however data looking at the influence of the broader URT microbiota in OM in indigenous populations will be lacking.

1.9.2 Aim 2: To identify health-associated bacterial species in the upper respiratory tract of Indigenous Australian children To date, microbial studies exploring the URT microbiota in relation to OM in Indigenous Australian children have focused on the three main otopathogens, using both culture-based analysis and species-specific PCR. Only one study has used 16S rRNA NGS in Indigenous Australian children, in a very small population with OME. Few studies have investigated the presence of respiratory viruses in this population. This thesis aims to use a culturomics approach, combined with culture- independent analysis (16S rRNA NGS, quantitative RT-PCR of otopathogen load and respiratory viruses) to explore the URT microbiota of a cohort of Indigenous Australian children. This will be the most comprehensive study of the URT microbiota in Indigenous Australian children to date. The microbial data will be analysed in relation to URT health, compared to OM/ URTI, and co- colonisation correlations to provide information on possible health-associated species.

Hypothesis: Healthy Indigenous Australian children will have different URT microbial profiles when compared to children with OM/ URTI. These profiles are more likely to be dominated by AHS and lactobacillus.

1.9.3 Aim 3: To explore the ability of health-associated bacterial species to inhibit the growth of otopathogens Aim 2 will provide a biobank of cultured isolates in which to explore the bacterial interference of health-associated bacterial strains from Indigenous Australian children against otopathogens from the same child/ same community. Within this aim there are two sub-aims:

Aim 3.1: To explore the ability of commensal flora from healthy Indigenous Australian children to inhibit the growth of otopathogens The objective of this aim is to find strains that will be candidates for probiotic development. This will consist of an initial mass screening using agar overlay assays followed by cell-free supernatant studies for promising strains.

Hypothesis: There will be health-associated species in the URT of healthy Indigenous Australian children with excellent ability to inhibit the growth of all three otopathogens.

Aim 3.2: To explore whether there is a difference between the ability of commensal flora from healthy children to inhibit otopathogens commensal flora isolated from children with CSOM

The objective of this sub-aim is to explore whether bacterial interference within the URT microbiota of the individual child contributes to the pathogenesis of OM. Otopathogens can colonise the URT of healthy children without causing disease. This sub-aim investigates whether the strength of inhibition of the commensal flora influences whether the colonised child develops OM or not.

Hypothesis: Commensal flora from healthy Indigenous Australian children will be more effective at inhibiting the growth of otopathogens while commensal flora from children with CSOM will have weak/ no inhibition against otopathogens.

1.9.4 Aim 4: To complete first stage testing for probiotic development for bacterial strains identified as potential probiotic candidates in Aim 3.1 Bacterial strains identified as having excellent ability to inhibit the growth of otopathogens in Aim 3.1 will undergo testing as outlined by the European Food Safety Authority. Within the timeline and funding limitations of the thesis this will include comprehensive identification of the strains, whole genome sequencing and analysis, and antimicrobial susceptibility.

Chapter 2: Systematic Review Citation: Coleman A, Wood A, Bialasiewicz S, Ware RS, Marsh R. L., & Cervin, A. (2018). The unsolved problem of otitis media in indigenous populations: A systematic review of upper respiratory and middle ear microbiology in indigenous children with otitis media. Microbiome; 6:1–15.

This chapter has been published in the journal Microbiome. The original manuscript has been reformatted for this thesis; the original paper is available at:

Contributor Statement of contribution Andrea Coleman Conceptualisation and design: 75% Data collection & extraction: 50% Analysis and interpretation: 75% Drafting and writing of manuscript: 75% Amanda Wood Conceptualisation and design: 5% Data collection & extraction: 50% Analysis and interpretation: 5% Drafting and writing of manuscript: 5% Seweryn Bialasiewicz Conceptualisation and design: 5% Analysis and interpretation: 5% Drafting and writing of manuscript: 5% Robert Ware Conceptualisation and design: 10% Analysis and interpretation: 10% Drafting and writing of manuscript: 5% Robyn Marsh Analysis and interpretation: 5% Drafting and writing of manuscript: 5% Anders Cervin Conceptualisation and design: 5% Analysis and interpretation: 5% Drafting and writing of manuscript: 5% https://microbiomejournal.biomedcentral.com/articles/10.1186/s40168-018-0577-2.

Microbiome research is rapidly evolving and as techniques in both the molecular and culture-based paradigms advance, more is being known about the complex microbial basis of infectious diseases. To adequately progress the knowledge base regarding OM in Indigenous Australian children, there needs to a comprehensive assessment of the literature regarding the role of both potential pathogens and commensal flora. In this chapter I address the Aim 1, to explore the microbial data relating to OM in indigenous populations globally. The scope of the systematic review stretches beyond Indigenous Australian children as most indigenous populations suffer from high rates of OM, often related to risk factors linked to socioeconomic disadvantage. Similarly, there has been little change in prevalence of OM in indigenous communities globally despite antibiotic use and vaccination programs. A greater understanding of the microbiota in relation to OM in indigenous populations may facilitate innovation in prevention and treatment. Systematic review of the evidence is the first step to consolidating what is already known and highlighting key deficits in the literature.

The unsolved problem of otitis media in indigenous populations: A systematic review of upper respiratory and middle ear microbiology in indigenous children with otitis media

Andrea Coleman1,2, Amanda Wood3, Seweryn Bialasiewicz2, Robert S Ware4, Robyn L Marsh5, Anders Cervin1,6.

Affiliations: 1Faculty of Medicine, The University of Queensland, Brisbane, Queensland, Australia 2Queensland Paediatric Infectious Disease Laboratory, Centre for Children’s Health Research, Children’s Health Queensland Hospital & Queensland University of Technology, Child Health Research Centre, The University of Queensland, Brisbane, Queensland, Australia 3The Deadly Ears Program, Children’s Health Queensland Hospital and Health Service, Brisbane, Queensland, Australia 4Menzies Health Institute Queensland, Griffith University, Brisbane, Queensland, Australia 5Menzies School of Health Research, Charles Darwin University, Darwin, Northern Territory, Australia 6Department of Otolaryngology Head & Neck Surgery, Royal Brisbane and Women’s Hospital, Brisbane, Queensland, Australia

Abstract Background: Otitis media (OM) imposes a great burden of disease in indigenous populations around the world, despite a variety of treatment and prevention programs. Improved understanding of the pathogenesis of OM in indigenous populations is required to advance treatment and reduce prevalence. We conducted a systematic review of the literature exploring upper airway and middle ear microbiota in relation to OM in indigenous children.

Methods: Papers targeting microbiota in relation to OM in children <18 years indigenous to Australia, New Zealand, North America, and Greenland were sought. MEDLINE, CINAHL, EMBASE, Cochrane Library, and Informit databases were searched using key words. Two independent reviewers screened titles, abstracts, and then full-text papers against inclusion criteria according to PRISMA guidelines.

Results: Twenty-five papers considering indigenous Australian, Alaskan and Greenlandic children were included. There were high rates of nasopharyngeal colonization with the three main otopathogens (Haemophilus influenzae, Streptococcus pneumoniae, and Moraxella catarrhalis) in indigenous children with OM. Middle ear samples had lower rates of otopathogen detection, although detection rates increased when molecular methods were used. Pseudomonas aeruginosa and Staphylococcus aureus were commonly detected in middle ear discharge of children with chronic suppurative OM. There was significant heterogeneity between studies, particularly in microbiological methods, which were largely limited to culture-based detection of the main otopathogens.

Conclusions: There are high rates of otopathogen colonization in indigenous children with OM. Chronic suppurative OM appears to be associated with a different microbial profile. Beyond the main otopathogens, the data are limited. Further research is required to explore the entire upper respiratory tract/ middle ear microbiota in relation to OM, with the inclusion of healthy indigenous peers as controls.

Keywords: otitis media, indigenous, microbiota, paediatrics, systematic review

Otitis media (OM) describes a spectrum of pathologies that involve inflammation and/or infection in the middle ear. This spectrum encompasses a continuum from acute to chronic disease that is clinically characterised by fluid in the middle ear (4-7). OM is highly prevalent in indigenous populations globally, particularly when compared to non-indigenous peers (9, 131), and often occurs earlier, more frequently and in more severe forms (7, 9, 10). Prevalence data reports that up to one-third of Greenlandic and Alaskan Inuit, Native American and Australian Indigenous children suffer from chronic suppurative OM (CSOM) (23, 131-134). The World Health Organization considers CSOM prevalence of ³4% indicative of a public health problem serious enough to require urgent attention (17). OM-related complications result in approximately 21,000 deaths each year worldwide (33). OM-associated hearing loss can impact significantly on language and social skills development, school attendance and educational outcomes, and downstream effects such as greater contact with the criminal justice system later in life (7, 35, 36). Medical interventions including liberal antibiotic prescription and vaccination programs have limited effectiveness in indigenous populations (14, 27, 32), thus new treatment avenues need to be considered.

The reasons for high OM prevalence in indigenous populations are likely to be multi-factorial. Risk factors include poverty, inadequate housing, overcrowding, and exposure to environmental tobacco smoke (11, 23, 131, 135). These risk factors are ubiquitous across indigenous populations worldwide (136). Genetic susceptibility to OM has not been studied in indigenous populations (137, 138).

We use the term microbiota to refer to the bacterial taxa reported for upper respiratory and middle ear samples, while “microbiome” refers to “the catalogue of these microbes and their genes” (139). The microbiota of the upper respiratory tract (URT) is an important OM risk factor across all populations. Most research to date has focused on the role of the three main otopathogens: Streptococcus pneumoniae, Moraxella catarrhalis, and non-typeable Haemophilus influenzae (61). It is not currently clear whether commensal bacteria amongst the URT microbiota contribute to, or mitigate, OM risk in indigenous children. In non-indigenous children, 16S ribosomal RNA (rRNA) gene analyses have suggested that a ‘healthy’ nasopharyngeal (NP) microbiota is more diverse than that of children with OM (3, 44, 77, 140). This ‘healthy’ NP microbiota contains bacteria that may be protective or promote microbiota stabilization, including, Moraxella, Corynebacterium,

Dolosigranulum, Propionibacterium (Cutibacterium), Lactococcus and Staphylococcus (3, 44, 77, 140). It is currently unknown whether these results are generalizable to indigenous populations. While high rates of OM are reported for many developing countries, indigenous populations, as defined by the United Nations (141), share unique challenges in relation to OM. Otitis media in indigenous populations is difficult to prevent and treat, therefore we need to gain a better understanding of the microbial pathogenesis to establish knowledge gaps, provide direction for future research and help guide appropriate prevention and treatment options. The aim of this systematic review is to assess the current knowledge regarding the microbiological aetiology of OM in indigenous children from around the world by examining data pertaining to upper respiratory and middle ear samples.

Methods used for this systematic review were developed with reference to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement. The protocol was registered with the International Prospective Register of Systematic Reviews (CRD42016033905) prior to commencement.

2.2.1 Inclusion criteria All studies exploring the microbiota of the URT (nose, nasopharynx, mouth, oropharynx, throat, tonsils, adenoid, and middle ear) in relation to OM in indigenous children aged 0-18 years old were included. For studies that included children without OM and/or did not report microbiology results specifically for children with OM, either only middle ear data were included, or if only the NP was sampled, the studies were excluded. Indigenous populations from Australia, New Zealand, United States of America, Canada, and Greenland were included.

2.2.2 Search strategy Literature search strategies were developed in collaboration with a health sciences librarian using medical subject headings (MeSH) and key words (Supplementary Data). The following electronic databases were searched from inception until 15 August 2017: MEDLINE (from 1946) and CINHAL (from 1982) via EBSCOhost, EMBASE (from 1966), Cochrane Library (from 1996), and Informit (from 1990- April). To ensure search saturation, we reviewed the reference lists of relevant studies and sought unpublished clinical audits through the Australian Institute of Health and Welfare (https://www.aihw.gov.au/) and The Australian Indigenous Health Info Net 65

(https://healthinfonet.ecu.edu.au/). Two independent reviewers (ACol and AW) revised titles and abstracts, then full text publications with reference to the inclusion criteria. Study selection inter- rater agreement between the two reviewers was calculated as the proportion of positive agreement (PA) (142).

2.2.3 Data extraction Two independent reviewers (ACol and AW) extracted data in duplicate onto a Microsoft Excel spreadsheet. Publication authors were contacted where data had been represented graphically or data were missing. We screened for multiple reports from the same study, and where multiple reports existed, compared and extracted relevant data; if inconsistencies existed, we contacted the authors for clarification. The following data were extracted for all studies meeting inclusion criteria: publication year, geographical location, study design, number of participants, age range, ethnicity, number of participants with an OM diagnosis, type of OM, number of controls, anatomical location of sample(s), microbiota investigation method, type and quantity of bacteria, viruses, and fungi detected from each anatomic site. For the purpose of the review ‘culture’ is defined as culture targeting the three main otopathogens and ‘extended culture’ is defined as culture used to detect bacteria beyond these otopathogens. Only quantitative PCR (qPCR) data were included when both culture and qPCR were used. For longitudinal studies, data relating to both the number of swabs and number of children were extracted, when there were multiple swabs per child. For data obtained from clinical trials, we included data only from samples collected prior to randomization.

2.2.4 Data analysis Where there were a sufficient number of studies, meta-analysis of proportions were calculated using random effects analysis via Stata/IC 15, otherwise we synthesized the data into a systematic narrative. We calculated heterogeneity using I2 statistic.

2.2.5 Risk of bias assessment Two independent reviewers (ACol and AW) assessed the risk of bias for each study with reference to the Critical Appraisal Skills Program (CASP) Cohort Study Checklist (143). Within the CASP Checklist, we assessed for the following confounding variables: age, overcrowding, antibiotic use, daycare/ school attendance, and concurrent respiratory/ upper respiratory tract infection. Study quality was categorized as 'poor', 'moderate' or 'good' based on the CASP Checklist. The overall quality of evidence was judged as high, moderate, low, and very low (144). 66

The initial search identified 5592 articles. After screening titles, the abstracts of 956 articles and 332 full text publications were reviewed (Figure 7). There was substantial PA between the reviewers of titles (PA = 0.68) and abstracts (PA = 0.79). Twenty-five articles met the inclusion criteria; these were from Australian Indigenous (n = 22), Greenlandic (n = 2) and Alaskan Inuit (n =

1). No papers reported OM otopathogens or microbiota in Native American or New Zealand Maori children.Figure 1: Literature search and selection

Records identified through database searching: 5592

Identification Did not meet inclusion criteria: 4636

Eligible articles for abstract review: 956

Screening Did not meet inclusion criteria: 624

Eligible articles for full text review: 332

Eligibility Did not meet inclusion criteria: 307

Articles included in review: 25

Australian Indigenous: 22; Greenland: 2; Alaskan Native: 1; Native American: 0; Maori: 0.

Included

AOM/AOMwP: 7 OME: 6 CSOM: 5 CSOM/AOMwP: 5 All types OM: 3

Figure 7: Literature search and selection

2.3.1 Risk of bias assessment According to the CASP risk of bias assessment, most studies (80%) were judged as either ‘poor’ or ‘moderate’, largely due to confounding variables not being considered (Table 6). Recruitment bias was difficult to assess, as recruitment processes were often poorly documented. Only one study (145) included healthy indigenous controls and another three (11, 57, 67) included children without

OM enrolled in longitudinal studies. Within indigenous populations, participants were recruited from limited geographical regions, making generalization beyond these regions difficult. Overall, the quality of the literature was ‘low’.

Table 6: Risk of bias assessment

focused issue? focused Cohort recruitment acceptable? Exposure accurately measured? Outcome accurately measured? factors confounding Important identified? factors confounding Important accounted for? Are the results precise? Are the results believable? Do results fit with other available data Overall quality score Reference Did the study address a clearly

1972, Stuart* (146) + + - - + - - - - Poor 1975, (147) ------? - - Poor Copeman 1975, Stuart (148) + + ------Poor 1985, (65) + - + + - - ? - - Poor Dawson 1994, Leach (11) + + + + - - - + + Mod 1996, Homøe (145) + + + + + + + + + Good 1999, (149) + + + + - - + + + Mod Parkinson 2003, (150) + ? + + + + + + + Good Couzos* 2003, Stuart (66) + ? + + + - + + + Mod 2005,Gibney (63) + + + ? - - + + + Poor 2006, Leach* (151) + ? + + - - - - + Poor 2007, (152) Ashhurst- + ? + + - - ? + - Mod Smith 2008, Leach* (153) + + + + - - + + + Mod

2008, Leach* (154) + + + + - - + + + Poor 2009, (87) + + + + - - - - - Poor Homøe* 2009, (31) + ? - + - - + - + Poor Mackenzie* 2010, Morris (155) + + + + + + + + + Good * 2011, Binks (57) + ? ? + - - + + + Poor 2012, Marsh (38) + ? + + - - + + + Mod 2012, Sun (67) + + + + + + + + + Good 2013, Smith- (64) + + + + - - + + + Mod Vaughan 2013, (156) + ? + + - - + + + Mod Stephen* 2015, Jervis- (1) + ? + + + + + + + Good Bardy 2015, Leach* (157) + ? + + - - + + + Mod 2016, Leach* (14) + ? + + + - + + + Mod Note: Data based on CASP-based risk of bias assessment. Assessment of bias pertained to the microbiology data, and not to clinical data. * Indicates studies where microbiological outcomes were not the primary outcome. ? indicates that this variable was unable to be assessed.

2.3.2 Heterogeneity The literature was limited by methodological and statistical heterogeneity across the studies, including heterogeneity in study design, participant age, OM diagnosis, and laboratory methods (Table 7) Where there were sufficient data to calculate I2; most were >70%, indicating moderate- high heterogeneity (Figure 8, Figure 9, Figure 10).

Table 7: Characteristics of included studies